Quality Assurance in Healthcare Service Delivery, Nursing and Personalized Medicine: Technologies and Processes

Quality Assurance in Healthcare Service Delivery, Nursing, and Personalized Medicine: Technologies and Processes Athina...

Author: Athina Lazakidou

56 downloads 992 Views 8MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Quality Assurance in Healthcare Service Delivery, Nursing, and Personalized Medicine: Technologies and Processes Athina Lazakidou University of Peloponnese, Greece Andriani Daskalaki Max Planck Institute for Molecular Genetics, Germany

Senior Editorial Director: Director of Book Publications: Editorial Director: Acquisitions Editor: Development Editor: Production Editor: Typesetters: Print Coordinator: Cover Design:

Kristin Klinger Julia Mosemann Lindsay Johnston Erika Carter Hannah Abelbeck Sean Woznicki Jennifer Romanchak Jamie Snavely Nick Newcomer

Published in the United States of America by Medical Information Science Reference (an imprint of IGI Global) 701 E. Chocolate Avenue Hershey PA 17033 Tel: 717-533-8845 Fax: 717-533-8661 E-mail: [email protected] Web site: http://www.igi-global.com Copyright © 2012 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher. Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.

Library of Congress Cataloging-in-Publication Data

Quality assurance in healthcare service delivery, nursing, and personalized medicine: technologies and processes / Athina Lazakidou and Andriani Daskalaki, editors. p. cm. Summary: “This book offers a framework for measuring quality of service in the healthcare industry as it pertains to nursing, with insight into how new technologies and the design of personalized medicine have improved quality of care and quality of life”--Provided by publisher. Includes bibliographical references and index. ISBN 978-1-61350-120-7 (hardcover) -- ISBN 978-1-61350-121-4 (ebook) -- ISBN 978-1-61350-122-1 (print & perpetual access) 1. Nursing--Quality control. 2. Nursing assessment. 3. Medical innovations. I. Lazakidou, Athina A., 1975- II. Daskalaki, Andriani, 1966RT42.Q35 2012 362.1068’4--dc23 2011031206

British Cataloguing in Publication Data A Cataloguing in Publication record for this book is available from the British Library. All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the authors, but not necessarily of the publisher.

Editorial Advisory Board Sia Barzela, Radboud University Nijmegen Medical Centre, The Netherlands Amit Chattopadhyay, National Institutes of Health (NIH), & National Institute of Dental and Craniofacial Research (NIDCR), USA Cathrin Dressler, Laser- und Medizin-Technologie GmbH, Germany Penny Emmanouil, Penteli’s Children Hospital, Greece Anastasia Kastania, Athens University of Economics and Business, Greece Melpomeni Lazakidou, Institute of Occupational Medicine, Austria Valko Petrov, Bulgarian Academy of Sciences, Bulgaria Konstantinos Siassiakos, University of Piraeus, Greece Christoph Wierling, Max-Planck-Institute for Molecular Genetics, Germany

List of Reviewers Anastasia Kastania, Αthens University of Economics and Business, Greece Andriani Daskalaki, Max Planck Institute for Molecular Genetics, Germany Evangelia Kopanaki, University of Piraeus, Greece Melpomeni Lazakidou, General Hospital Salzburg, Austria Konstantinos Siassiakos, Supreme Council for Civil Personnel Selection, Greece Ibrahiem El Emary, King Abdulaziz University, Saudi Arabia Athina Lazakidou, University of Peloponnese, Greece Evangelos Kotsifakos, University of Piraeus, Greece Sophia Kossida, Academy of Athens, Greece Stelios Zimeras, University of the Aegean, Greece

Table of Contents

Preface...................................................................................................................................................vii Section 1 Clinical Diagnostic Methods Chapter 1 From Non-Invasive Hemodynamic Measurements towards Patient-Specific Cardiovascular Diagnosis................................................................................................................................................. 1 Stefan Bernhard, Freie Universität Berlin, Germany Kristine Al Zoukra, Freie Universität Berlin, Germany Christof Schütte, Freie Universität Berlin, Germany Chapter 2 Personalized Experience Sharing of Cai’s TCM Gynecology............................................................... 26 Guo-Zheng Li, Tongji University, China Mingyu You, Tongji University, China Liaoyu Xu, Shanghai Literature Institute of TCM, China Suying Huang, Shanghai Literature Institute of TCM, China Section 2 Basic Research: A Bridge to Modern Medicine Chapter 3 Stem Cell-Based Personalized Medicine: From Disease Modeling to Clinical Applications............... 49 Alessandro Prigione, Max Planck Institute for Molecular Genetics, Germany Chapter 4 Bifurcation Analysis of a Model Accounting for the 14-3-3σ Signalling Compartmentalisation......... 61 Svetoslav Nikolov, Institute of Mechanics and Biomechanics, Bulgaria Julio Vera, University of Rostock, Germany Olaf Wolkenhauer, University of Rostock, Germany

Chapter 5 Engineering Gene Control Circuits with Allosteric Ribozymes in Human Cells as a Medicine of the Future............................................................................................................................................... 71 Robert Penchovsky, Sofia University “St. Kliment Ohridski”, Bulgaria Section 3 Medical Treatment and Research: Ethics and Applications Chapter 6 Ethical Guidelines for the Quality Assessment of Healthcare............................................................... 94 Amit Chattopadhyay, National Institutes of Health (NIDCR), USA Khushdeep Malhotra, National Institutes of Health (NIDCR), USA Sharmila Chatterjee, George Washington University, USA Chapter 7 Systems Biology Methodologies for the Understanding of Common Oncogenetic Mechanisms in Childhood Leukemic and Rhabdomyosarcoma Cells.......................................................................... 111 George I. Lambrou, University of Athens, Greece Maria Adamaki, University of Athens, Greece Eleftheria Koultouki, University of Athens, Greece Maria Moschovi, University of Athens, Greece Section 4 Healthcare Quality Assessment Chapter 8 Planning of a Specialized Pedagogic Environment and Defining Ethical Requirements in Educational Practice for Healthcare Quality................................................................................... 170 Vassiliki Ioannidi, University of Peloponnese, Greece Antonios K. Travlos, University of Peloponnese, Greece Sofia Vasileiadou, University of Athens “Attikon” Hospital, Greece Evangelia Loukidou, Loughborough University, UK Chapter 9 Τhe Social Role of Technology in Healthcare Quality Improvement.................................................. 178 Anastasia K. Kadda, University of the Peloponnese, Greece Chapter 10 Intelligent System to Quality Assurance in Drugs Delivery................................................................ 187 Antonio J. Jara, University of Murcia (UMU), Spain Mona Alsaedy, Kingston University London, UK Alberto F. Alcolea, University of Murcia (UMU), Spain Miguel Zamora, University of Murcia (UMU), Spain Antonio F. Gómez-Skarmeta, University of Murcia (UMU), Spain

Chapter 11 Quality in Telemedicine Services........................................................................................................ 203 Eleni Christodoulou, University of Peloponnese, Greece Sofia Zyga, University of Peloponnese, Greece Maria Athanasopoulou, University of Athens, Greece Chapter 12 Quality Based on a Spatial SERVQUAL Model in Healthcare........................................................... 209 Stelios Zimeras, University of the Aegean, Greece Chapter 13 Bayesian Ontologies in Spatial Integrating Medical Information Systems......................................... 220 Stelios Zimeras, University of the Aegean, Greece Chapter 14 Improving Quality of Care through Risk Management Knowledge Transfer and Quality Assurance in Nursing Care.................................................................................................................. 233 Kleopatra Alamantariotou, University Hospital of Larissa, Greece Chapter 15 Role of Information Technology in Healthcare Quality Assessment................................................... 242 Stamatia Ilioudi, University of Peloponnese, Greece Athina Lazakidou, University of Peloponnese, Greece Maria Tsironi, University of Peloponnese, Greece Compilation of References................................................................................................................ 254 About the Contributors..................................................................................................................... 279 Index.................................................................................................................................................... 288

vii

Preface

In general, outcome measurement has focused on a health gain or health maintenance score, or an overall wellbeing result. However, because quality of life is difficult to define and even more difficult to measure - particularly with physically and mentally vulnerable people - outcomes from nursing in continuing care are not easily articulated. The focus of the nursing assessment tool is therefore on increasing quality of life, rather than perceiving health gain simply as increased longevity. Assessment is considered to be the first step in the process of individualized nursing care. It provides information that is critical to the development of a plan of action that enhances personal health status. It also decreases the potential for, or the severity of, chronic conditions and helps the individual to gain control over their health through self-care. One of the great challenges in medicine is to deliver effective therapies tailored to each patient based on his molecular signatures. The so-called “personalized medicine” would involve tools of patient evaluation that would tell clinicians the correct drug and doses for the patient. Patient outcomes of a drug intervention are the result of conditional probabilistic interactions between complexes of drugmetabolizing enzyme genes, a range of metabolic regulatory genes, and environmental factors. Systems biology tools and concepts that integrate modelling of signalling pathways and regulatory networks at many levels of biological organization in the whole organism provide a help to personalized medicine. The main goal of this new book is to provide innovative and creative ideas for improving the quality of care and to explore all new technologies in medical informatics and health care delivery systems as well as technological advances in personalized medicine. The topics of this book cover useful areas of general knowledge including concepts of quality, quality assessment and quality assurance, risk management and quality of care, patient and nurse assessment in personalized medicine, role of information technology in healthcare quality assessment, quality in telemedicine services, tools and techniques in systems biology, ethical guidelines for the quality assessment of healthcare, health system policies, and service delivery. The book covers basic concepts, best practices, techniques, investigative challenges in clinic and research.

Organization of the Book The book is divided into four sections: Section One: “Clinical Diagnostic Methods” introduces innovative concepts in medical diagnostics. Chapter 1 presents patient specific cardiovascular diagnosis. The novel method described in this chapter produces a distribution of physiologically interpretable models that allow the identification of disease

viii

specific patterns that corresponded to clinical diagnoses, enabling a probabilistic assessment of human health condition. In this work a technique is presented to identify arterial stenosis and aneurisms from anomalous patterns in signal and parameter space. Chapter 2 introduces experience sharing of Cai’s gynecology in Traditional Chinese Medicine. A novel system named TCM-PMES is demonstrated. This system preserves the diagnostic processes in a personalized way. Section Two, “Basic Research: A Bridge to Modern Medicine,” serves as a comprehensive introduction to methods supporting basic research. The emphasis of chapter 3 is on new therapeutic approaches using functional cells derived from stem cells. This chapter provides a general overview on technologies applied on stem cell research. Chapter 4 gives an extended analytical consideration of mathematical modelling to the analysis of biological pathways perturbated in disease. Chapter 5 discusses the development of novel universal strategies for exogenous control of gene expression based on designer riboswitches that can function in the cell. Section Three is titled “Medical Treatment and Research: Ethics and Applications.” Chapter 6, entitled “Ethical Guidelines for the Quality Assessment of Healthcare,” describes how ethical principles can be used as guidelines for the quality assessment of healthcare provision. In Chapter 7, the authors present the computational workflow to investigate the process of tumor progression, and present this approach through an example of childhood neoplasias. Section Four, “Healthcare Quality Assessment,” is dedicated to methods applied in the healthcare service delivery. In chapter 8, the authors define ethical requirements in educational practice for healthcare quality. In chapter 9, the author describe the social role of technology in healthcare quality improvement. Chapter 10 presents a novel solution to improve quality assurance, in drugs delivery; i.e. reduce clinical errors caused by drug interaction and dose. Telemedicine services should meet the international quality requirements in order to accomplish quality assurance in healthcare provision. Technology advances have brought forward new and evolved services and technical infrastructure that promote and enhance quality healthcare services, such as telepresence and wearable technology. Nevertheless, there are several obstacles in telemedicine performance that need to be resolved. In Chapter 11 the authors try to define quality issues in Telemedicine Services. The main purpose of the Chapter 12 is to represent an alternative effective model for measuring the quality of healthcare (SERVQUAL) considering the geographical location of the under examination healthcare sectors. Geographical Information Systems (GIS) play a major role in all areas of health research, especially for the understanding of spatial variations concerning disease monitoring. Chapter 13 describes the methodological approach for the development of a real time electronic health record, for the statistical analysis of geographic information and graphical representation for disease monitoring. The purpose of the Chapter 14 is to provide innovative knowledge and creative ideas of improving quality of care and to explore how risk management and Knowledge transfer and quality assurance can improve health care. Under careful consideration, our purpose is to summarize which factors improve and promote the quality of care and which factors diminish quality. There are forms of ongoing effort to make performance better. Quality improvement must be a long- term continuous effort, reducing errors and providing a safe trust environment for health professionals and patients. After reading this chapter, the reader should know the answer to these questions: What role can risk management and knowledge transfer play in quality of care? How risk management and knowledge transfer can work together? What are the factors that improve risk management and quality assurance in health care? How does knowledge transfer support inform and improves care?

ix

Quantifying and improving the quality of health care is an increasingly important goal in medicine. Because quality of life is difficult to define and even more difficult to measure - particularly with physically and mentally vulnerable people - outcomes from nursing in continuing care are not easily articulated. The focus of the nursing assessment tool is therefore on increasing quality of life, rather than perceiving health gain simply as increased longevity. Assessment is considered to be the first step in the process of individualized nursing care. It provides information that is critical to the development of a plan of action that enhances personal health status. It also decreases the potential for, or the severity of, chronic conditions and helps the individual to gain control over their health through self-care. In the Chapter 15 the authors try to describe how important is the role of Information and especially of the Information Technology in Healthcare Quality Assessment. The book, “Quality Assurance in Healthcare Service Delivery, Nursing and Personalized Medicine: Technologies and Processes,” contains text information, but also a glossary of terms and definitions, contributions from international experts, in-depth analysis of issues, concepts, new trends, and advanced technologies in Healthcare Service Delivery, in modern clinical diagnostics, and in advanced medical research. The new book will be an excellent source of comprehensive knowledge and literature on the topic of quality assurance in healthcare service delivery, nursing and personalized medicine. All of us who worked on the book hope that readers will find it useful. Athina Lazakidou University of Peloponnese, Greece Andriani Daskalaki Max Planck Institute for Molecular Genetics, Germany

Section 1

Clinical Diagnostic Methods

1

Chapter 1

From Non-Invasive Hemodynamic Measurements towards Patient-Specific Cardiovascular Diagnosis Stefan Bernhard Freie Universität Berlin, Germany Kristine Al Zoukra Freie Universität Berlin, Germany Christof Schütte Freie Universität Berlin, Germany

ABSTRACT The past two decades have seen impressive success in medical technology, generating novel experimental data at an unexpected rate. However, current computational methods cannot sufficiently manage the data analysis for interpretation, so clinical application is hindered, and the benefit for the patient is still small. Even though numerous physiological models have been developed to describe complex dynamical mechanisms, their clinical application is limited, because parameterization is crucial, and most problems are ill-posed and do not have unique solutions. However, this information deficit is imminent to physiological data, because the measurement process always contains contamination like artifacts or noise and is limited by a finite measurement precision. The lack of information in hemodynamic data measured at the outlet of the left ventricle, for example, induces an infinite number of solutions to the hemodynamic inverse problem (possible vascular morphologies that can represent the hemodynamic conditions) (Quick, 2001). Within this work, we propose that, despite these problems, the assimilation of morphological constraints, and the usage of statistical prior knowledge from clinical observations, reveals diagnostically useful information. If the morphology of the vascular network, for example, is constrained by a set of time series measurements taken at specific places of the cardiovascular system, it is possible to solve the hemodynamic inverse problem by a carefully designed mathematical forward model in combination with a Bayesian inference technique. DOI: 10.4018/978-1-61350-120-7.ch001

Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.

From Non-Invasive Hemodynamic Measurements towards Patient-Specific Cardiovascular Diagnosis

The proposed cardiovascular system identification procedure allows us to deduce patient-specific information that can be used to diagnose a variety of cardiovascular diseases in an early state. In contrast to traditional inversion approaches, the novel method produces a distribution of physiologically interpretable models (patient-specific parameters and model states) that allow the identification of disease specific patterns that correspond to clinical diagnoses, enabling a probabilistic assessment of human health condition on the basis of a broad patient population. In the ongoing work we use this technique to identify arterial stenosis and aneurisms from anomalous patterns in signal and parameter space. The novel data mining procedure provides useful clinical information about the location of vascular defects like aneurisms and stenosis. We conclude that the Bayesian inference approach is able to solve the cardiovascular inverse problem and to interpret clinical data to allow a patient-specific model-based diagnosis of cardiovascular diseases. We think that the information-based approach provides a useful link between mathematical physiology and clinical diagnoses and that it will become constituent in the medical decision process in near future.

INTRODUCTION Diagnosis of cardiovascular diseases is an important and difficult task in medicine. Usually differential diagnostic approaches are applied, where several factors or symptoms are considered in an exclusion procedure. This procedure may lead to false assumptions that often result in misdiagnoses. Therefore, the attempt of using the knowledge and experience of experts collected in databases to support the diagnosis process by computational methods seems reasonable. This non-obvious task is valuable for decision making in today’s medicine. Typically various data mining frameworks are used to transform the data into therapeutically relevant knowledge for decision making. Several iterative subtasks process the data to discover the hidden information that diagnose diseases. Even though the amount and information content within clinical data has grown continuously and rapidly the number of accepted and reliable procedures for extracting diagnostically or therapeutically relevant information has not increased comparably. On the one hand the difficulty in the process of information extraction of richer data sets into improved clinical therapies lies in the traditional

2

deterministic view immanent in current mathematical models of physiological processes. On the other hand there is a lack of sufficient therapeutic options that can cope with advanced diagnostic information. The former problem traces back to the fact that current methods have limited ability to integrate the statistical nature of clinical data obtained from a broad patient population. Within this work we bridge the gap to a more suitable interpretation of model parameters and states observed in a broad patient population, namely by statistical analysis and interpretation. Such statistical models have been developed for a range of problems in biomedical signal analysis to remove contaminants from EEG and ECG signals (Sameni, 2008) and recently to estimate parameters of the cardiovascular system from blood pressure signals (McNames, 2008). It was shown that the extraction of information about the health condition of subgroups of the population with such methods provides the potential to develop new individualized therapeutic strategies that bring true benefit to the patients. Resolving this incongruity will pave the way for a statistical interpretation of cardiovascular system models that optimally use the information of specific subpopulations for diagnostic purposes.


The Intention of the Angiology Project Carried Out in the Innovation Laboratory “Mathematics for Diagnosis (M4D)” is twofold: the collection of physiologic information and associated the probabilistic models for the use in data base driven information systems and medical devices and to improve the understanding of human cardiovascular physiology. Within the project, there will therefore be many scientific subtasks concerning the parameterization of anatomical regions of the cardiovascular system, the data integration and mathematical model building. These include mechanical and hemodynamical aspects of the normal and diseased system. Some of the major components of the cardiovascular system for which computational models have been developed recently include the hemodynamic conditions in the coronary and systemic vascular circulation. Of particular interest are lumped-parameter models describing transport of blood accounting for the spatio-temporal distribution of blood and uptake of nutrients and oxygen within the vasculature. Clinical data has been essential in this area, both for estimating the structural and kinetic parameters of mass transport and for providing independent measurements that are used to validate such models. These models should provide an opportunity to answer important questions in integrative cardiovascular physiology that have abscond intuitive understanding. One example discussed in this study is the detection of vascular defects like stenosis or aneurisms that influence perfusion and cause acute ischemia. Following studies regarding the needs of clinical applications (Capova, 2003; Xiao, 2002; Wessel, 2000), we will have to combine physical and physiological aspects of pathological conditions in the cardiovascular system with patientspecific simulations based on non-invasively accessible data. Within our project we developed parametric computational techniques for model-

ing the mechanics, and transport processes, while incorporating kinetic and structural information from data measurements. In this approach a set of multi-transducer measurements, taken at several places of the vasculature, act as constraints to the hemodynamic inverse problem. Its solution allows to extract a variety of useful patient-specific diagnostic information like for example the network morphology including mechanical properties and the location of arterial stenosis and aneurisms, the stroke volume and ejection fraction of the heart and peripheral resistances. In the following we will discuss several mathematical models generally used to simulate the dynamics of the cardiovascular system and we will introduce an electrical analogue model developed within the M4D group. The model is used to show the success and failure in traditional parameter estimation procedures in example of an aortic aneurism. The failures lead us to a novel view of cardiovascular modeling where the hemodynamic inverse problem is stated as statistical inference problem. Finally we discuss the potential of the Bayesian inference technique in cardiovascular system identification and in the classification of diseased model states.

Mathematical Forward Models in Cardiovascular Physiology The application of mathematical models of physiological mechanisms could sufficiently enhance the interpretation of clinical data. However, such models are generally nonlinear and complex, so that the identification of parameters and states that represent available data are difficult to determine and interpret. Cardiovascular models have been widely used in physiology and medicine to support clinicians. Specifically, they provide insights on the behavior of blood pressures in the vasculature of the human circulatory system under normal and diseased conditions. Within this study we are especially interested in the modeling and detection

3


of pathologic defects like vascular stenosis and/ or aneurisms.

Common Cardiovascular System Models Several forward models of the cardiovascular system have been previously developed to describe and analyze the cardiovascular function on a theoretical basis. In one of the first theoretical investigations the arterial system was described by a Windkessel model, which reduces the system to a single elastic chamber that temporarily stores the blood (Stergiopulos, 1999). Although the benefits for calculation of cardiac output loomed large the model did not manage to describe the pulse wave propagation, because the Windkessel model cannot describe the spatial distribution of pressure and flow waves. This disadvantage is caused by the lumped description of the cardiovascular system as a single elastic chamber that cannot give a detailed description of all constituents (hundreds of arteries, arterioles and capillaries). The lumped description is however well satisfied and applicable as a boundary condition to model outflow conditions of the capillary bed (Quateroni, 2001). Today’s cardiovascular system models are based on the Navier-Stokes equation and the constitutive equation for the arterial wall or a set of equations resulting from the application of a variety of simplifying assumptions. For circular symmetric and linear elastic arteries blood flow is described by a set of nonlinear one-dimensional wave equations describing pressure and flow along the arteries (Womersley, 1955; Olufsen, 2000; Alastruey, 2009; Formaggia, 2009). Further reduction can be obtained by the long wave approximation, where the pressure variation along a vessel segment is assumed to be small, so that non-linearities can be removed. These equations can be solved by a semi-analytical approach if the vascular network is truncated after a few generations (Stergiopulos, 1992; Segers, 1998; Zamir, 2000; Pedley, 2008).

4

Anatomically detailed models describing propagation of pressure and flow waves by onedimensional equations are derived in several places (Olufsen, 2000; Formaggia, 2009; Alastruey, 2009). The authors of (Matthys, 2007) have used experimental models to verify and test parameters and hemodynamic conditions in one-dimensional model systems. The accuracy of the pressure and flow within their one-dimensional formulation was quantified to be 4% relative error in the pressure difference compared to the experiments. Further simplification of the one-dimensional system of partial differential equations (PDE) is obtained integrating out the spatial dependency. The resulting lumped system of zero-dimensional ordinary differential equations (ODE) (Olufsen, 2004) has similar structure as those used in electrical analog circuits. These models are often used as a Windkessel type boundary conditions in higher dimensional models (Quateroni, 2001), where they represent the dynamic peripheral impedance of the capillary bed (Stergiopulos, 1999), (Lee, 2004). However they also find application in several detailed descriptions of the coronary (Wang, 1989) and systemic (Westerhof, 1969; Avolio, 1980; John, 2004) vasculature. In Wang (1989) for example electrical analogue elements are used to model pressure and flow in 14 coronary arteries and in John (2004) a detailed representation of the whole systemic circulation is given. Both model systems where used to simulate the hemodynamics under normal and stenotic arterial conditions. In John (2004) the option of inverse calculation to determine the mechanical properties from blood flow and pressure waveforms is raised, but no further details are given.

The Cardiovascular System Model Used in the Study In the ongoing study we decided to set up a lumped parameter model of the cardiovascular system, because these models are known to reproduce the pulsatile pressure and flow in the systemic


vasculature quite realistically (Westerhof, 1969; Wang, 1989; John, 2004). The model is mathematically formulated in terms of an electric circuit analogue, where the electrical current is used as the analogue of the blood flow, F, and the electrical potential that of the pressure, P, in the circulatory system. Within the system of ordinary differential equations shown in Figure 1 (a) the pressure and flow in a vessel segment is modeled as RCL-circuit (three element Windkessel) consisting of a resistance, R, an inductor, L, in series as well as a capacity, C, in parallel (see Figure 1 (b)). The reader should be aware that the electrical circuit analogue is just for the sake of simple interpretation and graphical representation of the models to be discussed and does not mean that the cardiovascular system should be understood as a kind electrical circuit system. The pressure at the in- and outlet is denoted by Pin and Pout respectively, while the flow is denoted by Fin and Fout. The viscosity of the blood is modeled by a resistance, while the inductance comprises the inertia of the blood. The capacity in parallel models the linear elastic compliance of the arterial wall. Most electrical analogue models of the cardiovascular system are based on the pioneering work of Noordergraaf (1963). However, several modifications where done to adapt the model behavior (Stergiopulos, 1999; Hassani, 2007;

Alastruey, 2008). These comprise for example the inclusion of non-linearities or visco-elasticity of the vessel wall, derivation of elements to represent leakage flow, vascular defects or the inclusion of parameters that include additional vascular branches that are of particular interest in a study.

Parameterization of Cardiovascular Models In the forward problem a set of given model parameters, such as the length, cross-sectional area and elasticity, lead to the pressure and flow at specified locations in the vascular network. In practice it is often not possible to determine the cross-sectional area or elasticity for most of the arteries by direct measurement, even the segmental length is difficult to measure directly (Jovanovic, 2004). However good estimates where obtained in experimental studies, that computed the desired values from indirect measurements (Dobrin, 1978; Tardy, 1999). The parameter values found are widely spread and properties significantly change with the health condition and age of the subject (Zanchi, 1997; Hoeks, 1999). The strong variance within these parameters makes it difficult to simulate patient-specific settings, because the measurement of values for individual patients is time-consuming (Manor, 1994).

Figure 1. (a) ODE system describing the pressure and flow within the three element Windkessel model shown in (b)

5


The model parameters used in this work characterize a system of 242 first order ordinary differential equations like the ones given in Figure 1 (a), that are coupled to represent the morphology of the human arterial tree as given in (Westerhof, 1969). Within the model the blood flow in the arterioles and capillaries is lumped into a peripheral resistance, Z, that characterizes the pressure-flow relation at the network outlets. Parameters for the peripheral resistances where taken from Matthys (2007)

Numerical Solution of Cardiovascular Models The solution of the high-dimensional ODE system is conducted by common numerical solvers and allows us to predict the blood pressure and flow at arbitrary nodal points in the circulatory system. Within this study we are especially interested in the pressures at non-invasively accessible locations like the arteria radialis, the arterial carotis or the arteria brachials, because the signals can be directly compared with in-vivo pressure measurements to fit the model equations. In the next section we employ traditional optimization methods to re-estimate a set of parameters from pressure and flow waves that have been generated from the model outputs. Taking the simulated waves as observation data, we are able test a variety of hypothesis about vascular defects to dis-/prove the existence of specific diseases. In the adjacent section, we propose a statistical method that applies hypothesis based on a disease specific subpopulation as prior information to classify the measurement data automatically.

Parameter Estimation and Inverse Modeling in Cardiovascular Physiology The idea to use mathematical models to gain a better insight into the mechanisms of cardiovascular diseases, is not new but previous investiga-

6

tions concentrated on the solution of the forward problem (Westerhof, 1969; Olufsen, 2000). These models determine the blood and pressure distribution within the vasculature from specified structural and hemodynamical parameters to compare the results with in-vivo measurements. Inverse solutions have been seeked (Quick, 2001) but current methods only determine basic parameters like the total arterial compliance or the total peripheral resistance with sufficient accuracy. The general setting for model based data inversion in cardiovascular physiology is shown in Figure 2. Conventionally, reduced arterial system models are fitted to data, and the resulting model parameters are assumed to represent pretended arterial properties. The achievable accuracy of the parameters depends critically the quality of the data and complexity of the model used. Basic parameters like the total arterial compliance are estimated with high accuracy using simple models, while highly specific parameters like the location of vascular defects may even not be estimated using complex models, because the amount of data available is insufficient. This suggests a multi-model setting in which several models of different complexity are used to solve the hemodynamic inverse problem for different levels of detail. This implies that detailed models are lumped into less detailed macromodels to obtain the basic parameters, while the highly specific information is obtained by the detailed models.

Success and Failures of Current Parameter Estimation Methods in Cardiovascular Medicine In this section we will review some of the parameter estimation methods used in cardiovascular physiology. Subsequently we will resume what in our opinion is the main problem in cardiovascular parameter estimation. In Lee (2004) an electrical analog model for the aorta was investigated, where the authors optimized the parameters of four electrical circuit


Figure 2. Schematic overview of the model based hemodynamic inverse problem

models to fit transfer functions calculated from invasively measured pressure waves taken at the end of the aortic segments (Lathem, 1985). They estimated all electrical parameters with high accuracy and the output response generated at the end element was almost identical with the measured wave. Other methods try to reduce the order of the optimization problem to ensure the well-conditioning of the parameter estimation problem. In Samar (2005) a subset selection is therefore employed on a basic cardiovascular model consisting of left and right ventricles, systemic arteries and veins and pulmonary arteries and veins. A nonlinear least squares method was applied to identify the parameters. It was concluded, that the reduction of the parameter identification problem significantly improves the reliability of the estimates. Also in Heldt (2001) a subset of the entire parameter vector is determined by the Gauss-Newton method, matching simulated and measured data for a basic circulatory system. The authors of Thore (2008) demonstrated the applicability of a detailed one-dimensional wave propagation approach in combination with a traditional parameter estimation method to

identify patient-specific parameters. From a series of measurements they were able to reproduce the pressure wave in the abdominal aorta with high accuracy. John (2006) developed an inverse electrical circuit problem for the lower limb, that was solved by a globally convergent modification of the Newton’s method. Within the study the author’s also examined pathological changes like the detection of stenosis where the radius was determined sufficiently accurate. However they do not explain the exact procedure. Smith (2003) uses a six chamber closed loop model trying to simulate the cardiovascular system of a patient as accurately as possible. After detection of initial parameter values by trial and error parameters are approximated by unconstrained nonlinear optimization. Due to the description of the cardiovascular system as analog circuits, parametric estimation methods used to determine the structure and parameters of electrical circuits are of interest. Within Timmer (2000) and Korovkin (2007) structural properties and parameters for electrical elements where reliably estimated from signal measurements.

7


From the above survey it becomes clear that, even though numerous investigators were able to determine selected parameters from measurements, a detailed description of methods that are able to determine parameters for the whole circulatory system are not available until now. We assume that this deficit is majorly provoked by insufficient data mining methods currently used in cardiovascular system identification.

Parameter Estimation Example in a Cardiovascular System with Aortic Aneurisms Within this section we seek a set of parameters that represent a certain characteristic pressure or flow distribution measured for a specific disease. We imply that observable changes in the pressure distribution contain relevant information about a patient’s health condition. However for true measurements the parameters are unknown and cannot be used to validate the method, thus we generated three data sets from the model outputs (with known parameters) that are used as ideally observed data throughout the section.

Data Generation and General Simulation Setup Besides the normal parameterization of the systemic circulation model described in Westerhof (1969), we modified two aortic segments, (see segments 61 and 72 in Figure 8), to model either an thoracic aortic aneurism (TAA) or an abdominal aortic aneurism (AAA) in an ideal patient, respectively. The detailed parameters are given in Table 1. The network structure that corresponds to the numbering of the segments is shown in Figure 8. The simulated data sets comprise five pressure waveforms taken at the arteria vertebralis #8, arteria brachialis #14, arteria ulnaris #22, arteria femoralis #100 and arteria tibialis posterior #110 (see Figure 3 (a-c)). The results can

8

be compared to pressure waves for the AAA obtained in (Segers, 2009). To assure that the system is in steady-state the simulation duration was chosen to be 25 seconds, but only the time series of the final two seconds were used in the optimization.

Optimization Setup Due to the large number of variables we decided to use two constraint optimization algorithms: (i) a weighted variant of Levenberg-Marquardt nonlinear least squares algorithm that uses parameter sensitivities to control the step size (SENSOP) (Chan, 1999) and (ii) a non-linear steepest-descent algorithm (NLSD). For moderately sized models (of a few hundred parameters), the SENSOP is much faster than steepest-descent methods in a wide variety of problems. Even though the steepest descent algorithm is much slower, the implementation is simpler and the low storage requirement is advantageous for large problems. Finally, in some cases we employed a global optimization simulated annealing algorithm (SA). The upper and lower constraints for optimized parameters are set to physically expectable values. These are Clb=0 and Cub=3 for the lower and upper bound of the compliance and Llb, Rlb=0 and Lli, Rli=0.001 for the inductance and resistance respectively. The termination condition for the optimization is quantified by a threshold value for the root mean square deviation RMStr=0.001, computed for all data points in the simulated and measured time series.

Application of a Prior Hypothesis to Prove the Existence of Diseases The detection of diseases in this section is either based on a prior hypothesis (i) applying expected parameters for a specific disease as initial condition or (ii) on the assumption of a defect location. In (i) a set of expected parameters for a particular


Table 1. Modified parameter set from Westerhof (1969), for the generation of idealized pressure waves under normal and diseased conditions in an ideal TAA and AAA (changes are underlined) Normal

61

62

63

64

69

71

72

 ml     mmHg 

6.95E-02

7.96E-02

3.35E-02

3.00E-02

2.72E-02

2.41E-02

2.11E-02

L

 mmHg 2   s   ml 

9.00E-04

1.30E-03

2.90E-03

3.10E-03

3.50E-03

4.00E-03

4.50E-03

R

 mmHg   s  ml 

2.00E-04

3.00E-04

1.20E-03

1.70E-03

2.20E-03

2.70E-03

3.40E-03

61

62

63

64

69

71

72

 ml     mmHg 

2.78E+00

7.96E-02

3.35E-02

3.00E-02

2.72E-02

2.41E-02

2.11E-02

L

 mmHg 2   s   ml 

2.25E-04

1.30E-03

2.90E-03

3.10E-03

3.50E-03

4.00E-03

4.50E-03

R

 mmHg   s  ml 

1.25E-05

3.00E-04

1.20E-03

1.70E-03

2.20E-03

2.70E-03

3.40E-03

61

62

63

64

69

71

72

 ml     mmHg 

6.95E-02

7.96E-02

3.35E-02

3.00E-02

2.72E-02

2.41E-02

8.43E-01

L

 mmHg 2   s   ml 

9.00E-04

1.30E-03

2.90E-03

3.10E-03

3.50E-03

4.00E-03

1.10E-03

R

 mmHg   s  ml 

2.00E-04

3.00E-04

1.20E-03

1.70E-03

2.20E-03

2.70E-03

2.13E-03

C

TAA

C

AAA

C

disease is used to test the correlation to the measured data and in (ii) a set of normal parameters is used and the convergence of the optimization process in the assumed defect location is observed. Both approaches can be seen in analogy to the statistical classification procedure explained in the section “Future challenges”, except that the choice of locations and parameters are not single valued and entered by the user, but are expressed by prior probability distributions sampled from

a subpopulation of patients affected by the supposed disease. In the following examples we set up a series of hypothesis concerning the non-/existence of TAA, AAA and normal health conditions. The optimization results give hints that may lead us to a verification/falsification of the hypothesis. A. Allow parameters within one segment to be optimized, use normal starting parameters,

9


Figure 3. (a-c) Simulated ideal pressure waves for the five specified locations under normal (top) and diseased conditions for the TAA (middle) and AAA (bottom). Parameter modification is given in Table 1.

observe convergence and modified parameters to verify/falsify location hypothesis. B. Allow parameters in several locations to be optimized within a specific range, use normal parameters as initial condition, observe convergence and modified parameters to generate hint for location and kind of the disease. C. Assume diseased parameters as initial condition for a series of locations, observe initial correlation to verify/falsify diagnostic hypothesis. The above hypotheses can be combined and repeated to test several diseases where good parameter estimates already exist. This methodology can be regarded as manual cardiovascular screening by comparing the measurements to simulations generated by parameter sets that are related to specific diagnoses.

RESULTS AND PROBLEMS OF THE PARAMETER ESTIMATION Within this paragraph we show the results obtained by a variety of hypothesis to the measured data. Firstly we assume the defect within several locations of the network, i.e. either (a) single set of RCL parameters in segment 61, 62, 69, 71 or 72 is optimized, while the other segment parameters are constraint or (b) the parameters of multiple segments are optimized together. Finally (c) we apply a set of aneurism parameters to several segments 61, 62, 69, 71 and 72 to compare the correlation to the observed data generated for the TAA by RMS error. A. The optimization results indicate that obviously the best correlation with RMS of 9.77 E-5 is obtained if only segment 61 is optimized, while all other parameters are held fixed. In contrast the hypothesis that the defect is in 62 or 72 yielded only a RMS of

10


Figure 4. Comparison of the original pressure wave for TAA taken at the arteria femoralis (segment #100) to optimization results for selected hypothesis given in Table 2. The optimized data curve for a defect hypothesis in segment 61 is congruent with the original TAA data, while for a defect hypothesis in segment 72 the optimization yields decreasing correlation.

Table 2. Optimization for single segment hypothesis with SENSOP Hypothesis

estimated C

estimated L

estimated R

RMS

Num. of iterations

61

2.78E+00

2.25E-04

1.24E-05

9.77E-05

89

62

1.58E+00

3.68E-05

0.00E+00

4.19E+00

84, no success

72

2.11E-02

4.50E-03

2.28E-03

1.01E+01

26, no success

4.19 and 10 respectively (see Figure 4 and Table 2). This indicates that the aneurism is in segment 61. B. The results and parameters for the optimization of the compliance in segments 61, 62, 69, 71 and 72, indicate that an aneurism is located in segment 61, because the value for C61 is increased and close to the true value (see Figure 5 and Table 3) C. The hypothesis that a set of chosen parameters for an aneurism are true/false for segments 61, 62, 69, 71 and 72 is done by

comparing the RMS between the proposed locations. The RMS of 4.43 clearly indicates that the aneurism is in segment 61, because all other locations have increased RMS error (see Figure 6 and Table 4). Table 2 to 4 show the numerical results for the optimizations obtained for measurement data at locations #8, #14, # 22, #100 and #110 with hypothesis according to categories defined above. The number of iterations, RMS Error, true/estimated values are given to conclude the results.

11


Figure 5. According to the parameters in Table 3 the evidence that the defect is in segment 61 is assisted, because in either optimization over several parameters the values indicate that segment 61 has increased compliance.

Table 3A. Optimization for multiple segment hypothesis (C61-C72 as free parameters) with NLSD after 484 iterations and RMS of 3.90. It is clearly seen that the value of C61 is increased and approximates the true value. This is a hint for an aneurism in segment 61. Free parameter

Estimated Values

True Values

C61

1.75E+00

2.78E+00

C62

5.34E-02

7.96E-02

C69

6.20E-07

2.72E-02

C71

5.90E-02

2.41E-02

C72

5.47E-06

2.11E-02

Table 3B. Another hypothesis optimizing the RCL parameters of 61 and 72 yielded the same conclusion, because the NLSD found the true parameters with a RMS of 8.198E-4 in the signals. This indicates that the aneurism is more likely in segment 61.

12

Hypothesis

estimated C

estimated L

estimated R

RMS

Num. of iterations

61 / 72

2.778 / 2.104E-02

2.250E-04 / 4.499E03

1.253E-05 / 3.394E03

8.198E-04

1269, success


Figure 6. Comparison of the original pressure wave for TAA taken at the arteria femoralis (segment #100) to results for selected hypotheses about the location of the defect given a set of diseased parameters (see Table 4). The data curve for a defect hypothesis in segment 61 is most likely, because the RMS is smallest, while for segment 62, 69, 71 and 72 the RMS is increased.

Table 4. RMS error obtained with diseased parameter hypothesis of C=2.0, L=3E-4 and R=1.5E-4 applied to segments 61, 62, 69, 71 and 72. The RMS error indicates that the defect is in segment 61. Hypothesis

RMS error

61

4.43

62

9.10

69

10.89

71

10.40

72

10.20

Lessons Learnt so Far From the above results is becomes clear that, even though, optimization over all parameters is not feasible, a series of well chosen hypothesis leads to a conclusion revealing the defect in segment 61. However these results cannot be quantified into a one-to-one relation between defect and the parameters, but by a probabilistic relationship.

The above examples are based on ideal data, i.e. complete time series, no noise, no variability of patients or other influencing conditions. For true measurements the information content is much lower and the above parameter estimation will lead to vague estimates that cannot be used for diagnosis. However this putative information deficit generally causing the ill-posedness of the inversion problem can be overcome by statisti-

13


cal methods that use prior assumptions observed within a broad patient population. A theoretical basis for the solution of such problems is given in the framework of Bayesian statistics, which we discuss in the next section.

The Hemodynamic Inverse Problem Expressed as Bayesian Inference Problem Like in many fields of biological sciences and medicine, the methods that process signals obtained from measurements in cardiovascular physiology have to deal with uncertainty to obtain suitable results. The reason for the ill-posedness of model systems applied is manifold: (i) the number of measurements is limited, (ii) the measurements are imprecise, because they contain noise and artifacts and (iii) confounding variables (breathing, movement,...) are difficult to model and control. Indeed, each of these sources of uncertainty can be included in the data mining, if sufficient models for the underlying physiological phenomena exist and if the experimental conditions involved in the measuring process are known. In other words, problems in cardiovascular physiology are statistical in nature and their solution depends essentially on the quality of the data and the data mining routines used.

From Traditional Data Inversion to Statistical Inference Problems Traditional data inversion techniques have shown their success in well-posed deterministic problems that appear in many areas of sciences. Ill-posedness is conventionally, treated by regularization methods that improve the condition of the problem by more or less ad-hoc assumptions. Examples of common regularization methods are the truncated singular value decomposition, Tikhonov regularization or iterative methods like the Landweber-Friedman or Kaczmarz iteration and Krylov subspace methods (Kaipio, 2005). The

14

basic idea behind the regularization is to transform the ill-posed problem such that a unique solution can be obtained by standard methods that are robust to small errors in the data (Kaipio, 2005). However these methods mostly exclude important information from the solution procedure, thus the result is degenerated and may not represent important features contained in the data. Furthermore the representation of patient’s health condition by a single set of parameters is not suitable for diagnostic purposes, because it cannot describe the variances within a patient population. In contrast to traditional deterministic methods the objective of statistical inversion methods is to assess all information contained in the variables, based on the model and the measurement process, to express the uncertainty in terms of probability distributions. The solution to the inverse problem is then obtained by the evaluation of the posterior probability distribution. In other words, instead of a single estimate for the unknowns, as obtained by traditional inversion methods, the statistical method produces a continuous range of values to which the distributions assign a probability density value for each realization. In the setting of statistical inversion theory the ill-posedness is thus removed by restating the inverse problem as a well-posed extension in a larger space of probability distributions (Kaipio, 2005). We note that a delta function as posterior distribution in the statistical setting is equivalent to a single estimate in traditional inversion theory. The statistical basis of inversion techniques is given by Bayes formula, the problem is referred to as Bayesian inference problem. Bayesian statistical inference techniques have been used in several areas to extract useful information contained in physiological data sets (D’Argenio, 2006; Gerwinn, 2010; Hallander, 2010). In most applications concerning cardiovascular physiology the physical quantities and hence the measured data are functions of time and parameter estimation has to occur continuously. Such classes of inverse


problems are referred to as non-stationary inverse problems (Kaipio, 2005).

Non-Stationary Inverse Problems Bayesian recursive approaches rely on a statespace formulation of the inverse problem and provide useful tools to give continuous parameter estimates. The most popular representative is the extended Kalman filter (EKF) that can simultaneously estimate the parameters and unobserved states from noisy data of nonlinear time-continuous systems (Sitz, 2002). In contrast to the full Bayesian formulation it is restricted to uni-modal probability density functions, non-linear transition and measurement processes and Gaussian measurement noise. Recently (Zhang, 2004; McCombie, 2005; Swamy, 2007; Mukkamala, 2008) have proposed KF based methods to estimate the central aortic pressure waveform form two non-invasive peripheral pressure measurements. These methods are currently limited, because they cannot identify the pressure distribution within the rest of the vasculature. This deficit was discussed in (Thore, 2008) where the authors demonstrate the applicability of a one-dimensional wave propagation approach in combination with a traditional parameter estimation method to identify patient-specific parameters. Additionally to the achievements in (Zhang, 2004; McCombie, 2005; Swamy, 2007; Mukkamala, 2008), they where able to reproduce the abdominal aortic pressure waveform from non-invasive measurements with high accuracy. However, their method does neither include uncertainty inherent to measurements nor it is able to describe the variability of parameters found in a patient population. The subjective of the following paragraph is to include these deficits into a framework that allows to combine the variety of cardiovascular forward models with statistical inference methods for parameter estimation. The novel approach allows us to model patient-specific dynamics of the

cardiovascular system based on measurements and to extract characteristic waveforms and parameters that are statistically related to diseases that can be used for therapeutic purposes.

Problem Abstraction of the Multi-Measurement Setup The constraint hemodynamic inverse problem can be described as blind channel estimation problem with a single input (pressure-flow wave ejected by the left ventricle) and multiple outputs (multitransducer measurement) (SIMO). This problem is well known from wireless network communication problems (Tong, 1998) and has been successfully applied to problems in physiology to remove noise and artifacts from measurements (Sameni, 2006) or to estimate the central aortic pressure (Zhang, 2004; McCombie, 2005; Swamy, 2007; Mukkamala, 2008). In contrast to the classical input–output problem, where both input signal, s, and observed data, d, are available to estimate the transfer-function, the SIMO problem only relies on the observation signal, d, to identify both the input signal, s, and the parameters, λ, of the channels Hk. In Figure 7 we show a schematic of the SIMO estimation problem derived for the multi-measurement setup of the constraint hemodynamic inverse problem. The K different channels, Hk, with common source are obtained by the definition of K measurement locations. Each channel contains a forward map Ak, that relates the unknown parameters 𝜆k to the ideal data d (see Fig. 8). We assume that the measurement data is incomplete and contaminated by noise, so that the measured signals are assumed to have the form dk* = ξk Ak (λ) + ηk .

(1)

Here we denote the data loss by 𝜉k and the measurement noise by 𝜂k. We assume an additive white Gaussian noise (AWGN) model where

15


Figure 7. Statistical inversion scheme for the single input – multiple output problem arising by reformulation of the hemodynamic inverse problem as statistical inference problem.

Figure 8. Deterministic forward map for the channel left ventricle – arteria femoralis. In this case the physical model is given by a series of RCL-circuits, where we have assumed insignificant data loss and noise.

ηk  N (µk , σk 2 ) with mean 𝜇k and standard deviation 𝜎k. In the underlying formulation the forward maps may contain either RCL-circuits, transmission line or one-dimensional wave propagation elements describing a subsystem of the vasculature.

16

An example illustration for the channel from the heart to the arteria femoralis is given in Figure 8. The parameters for the channels and properties of the input can be estimated by time series analysis of the outputs, if the input probability distribution and moments are known.


Problem Formulation in Terms of Bayesian Inference The basis of statistical inversion techniques is given by Bayes formula πpost (λ | d * ) =

πpr (λ)π(d * | λ) π(d * )

,

(2)

that describes the statistical inference problem in terms of probability densities. πpr(λ) is the prior probability density for the parameters λ. It expresses what is known about the unknown parameter vector prior to the measurement, i.e. it contains prior information that does not depend on the actual measurement data d*. The posterior probability density πpost(λ|d*) is the conditional probability density of λ, given the observed data d*. The so called measurement likelihood, π(d*|λ), is the conditional probability of d* given λ. It expresses the likelihood of different measurement outcomes given d*. Finally the marginal probability π(d*) of d* acts as normalization constant. The value indicates if the model is consistent with the experimental observation. The solution to the statistical inverse problem is obtained by the determination of the parameters λ k given a set of measurements d*. It comprises three steps: (i) estimation of the prior probability density according to all prior information of λ, (ii) evaluation of the set of parameters where the measurement likelihood is maximal and (iii) computation of the posterior probability density.

Prior Modeling Within the theory of statistical inversion the available prior information is sufficiently included in the inversion process. The prior knowledge helps to formulate the ill-posed problem as a well-posed inferential problem. In cardiovascular

physiology such prior information consists of knowledge about the distributions of structural and biomechanical properties of the vasculature, like the vessel diameters, elasticities and lengths, but also the anatomic variants of the morphology of the network that is available prior to the measurement. The determination of appropriate model priors is challenging and consequently a variety different approaches have been introduced (Kaipio, 2005). In contrast to the specification of regularization functionals with either physical or ad-hoc assumptions, the construction of prior models involves the estimation of a probability distribution πpr(λ), that relates all values of the unknown λ to a probability value. Thus, values which are physically inadequate have a lower probability to appear than expectable ones. In the Bayesian approach model assumptions are typically proven by drawing several samples from the prior distribution and evaluating if the results are reasonable. Initially a Gaussian distribution is appropriate as prior model, because the mean and standard deviation for the parameters λ can be estimated from biomechanical experiments found in literature (Dobrin, 1978; Manor, 1994; Zanchi, 1997; Hoeks, 1999). The structural priors that describe the morphology of the vasculature can be estimated from experimental studies that explore anatomical variants (Jovanovic, 2004). These initial prior models however are continuously improved by clinical measurements or simulation data obtained for a broad patient population. Furthermore the definition of various samplebased priors, containing information about specific classes of diseases, serves as the basis for the design of a Bayesian classification algorithm (see later Section ÑFuture challenges“). Within the classification procedure these priors are referred to as representative ensemble that randomly generates samples that act as hypothesis to prove the existence of specific diseases.

17


Measurement Likelihood The measurement likelihood expresses the distribution of the measurement signals if all parameters λ are known. It conversely can be used to generate simulated signals for a known parameter set that can be compared to the actual measurements. Within our problem the likelihood can be defined as I

K

L(d 1*:I ,1:K | λ1:K ) = ∏ ∏ P[di*,k | λk ],

(3)

λ* = arg max L(d * | λ) .

(4)

i =1 k =1

λ

In other words, without assumptions about the parameter to estimate, one chooses the value that makes the output most likely. Within the context of Bayesian statistics this maximization problem can be solved even if the data is high-dimensional, incomplete and noisy. The general procedure is the expectation maximization (EM) algorithm discussed in the next paragraph.

Expectation Maximization Algorithm Herman O. Hartley (Hartley, 1958) pioneered the research on the EM algorithm in the late 1950s, the mathematical basis was given by Dempster, Laird, and Rubin (Dempster, 1977) in the late 1970s. Over the years, the EM algorithm has found many applications in various domains and has become a powerful estimation tool (McLachlan, 1997; Duda, 2001). In order to explain the EM procedure we will follow Neal and Hinton (Neal, 1993; Neal, 1998; Minka, 1998) who presented the EM algorithm from the perspective of lowerbound maximization. In order to apply MLE to our case, we formally assume a certain joint probability density function (PDF) π(d*,s|λ) between the observed and hidden random variables, and then marginalize

18

over the hidden random variable s, i.e, we have to integrate the joint PDF over the hidden state space Sh of s. This is the key idea behind the EM algorithm. According to this strategy, we compute the optimal parameter estimates as optimizers of the marginalized log-likelihood, logL.   λ* = arg max log L(d * | λ) = arg max log  ∫ p(d * , s | λ)ds    Sh λ λ

(5)

In the case of a discrete, finite hidden state space the integral in (5) has to replaced by a sum. The EM algorithm is an iterative optimization procedure. Starting with an initial parameter estimate λ (0), it is guaranteed to monotonically increase the likelihood function and converge towards some maximum of the Likelihood function. This maximum typically is the gloabal one, if some good initial estimate λ(0) is available. In our case, the initial estimate can be taken from the classical parameter estimation procedure.

Bayesian Classification as Concept in Patient-Specific Cardiovascular System Diagnosis Bayesian data integration addresses the classification problem by learning the distribution of instances of specific diseases in a subpopulation. The basic idea of Bayesian data integration and classification is discussed in (Firedman, 2002; Ceci, 2003). Within this study we use a naive Bayesian classifier, since more complex models may reduce the performance of the learning and inference processes. In order to outline the classification problem in the setting studied here, we have to start with the assumption that we have available the patientspecific measured signals d*i:I,k:K for M different patients that in some appropriate sense form a sampling of the patient population. Let us write d1*,Ö,dM*for the sequence of these measurements. Classification now means that we want to find out


which of the M patients belong to certain classes C1,Ö,CL. These classes can be, e.g., the class of patients that are do not have any stenosis, C0, and the classes of patients that have stenosis in segment number k of our model, Ck, which we will refer to as class k in the following. What we want to know is the probability that, conditional to the sequence of observed signals, patient j is in class k, i.e., we want to know P ( j ∈ C k | d1* , ..., dM* ) .

(6)

By using Bayes’ formula again, this probability can be computed from P ( j ∈ C k | d1* , ..., dM* ) =

P d1* , ..., dM* | j ∈ C k  P [ j ∈ C k ]

∑ P d1* , ..., dM* | j ∈ C k  P [ j ∈ C k ] K

(7)

where P[j∈Ck] denotes the prior probability of patients of class k and P[d1*,Ö,dM*|j∈Ck] denotes the probability of measuring the signal from a patient of class k. While the former probability is a classical prior with which we have to deal as described above, the latter probability has to be estimated algorithmically from the sequence of observations d1*,Ö,dM*|. Thus an appropriate classification algorithm has to perform two different estimations simultaneously: 1. Density estimation: Estimate the probability [d1*,Ö,dM*|j∈Ck] that a certain form of signal is observed from a patient of class k. 2. Classification: Find the hidden information whether patient j, from which signal dj* originates, belongs to class k. The combination of these two tasks in the sense of a joint likelihood optimization again leads to the EM algorithm. That is, in every step of the EM iteration first the density estimation is updated and then the classification probabilities. The iteration again converges if we chose appropriate initial

values and results in the optimal densities and classifications based on the available observation. Consequently, the results become better and better when more and more signals are added to the available information.

FUTURE CHALLENGES The aim of the above framework is to provide the basic tools that enable classification of measured data into classes with common properties – i.e. that are related to specific diseases. These classes can be used to classify unknown data measured at patients with unknown diagnosis by means of fuzzy probabilities. Such strategies have been applied in several areas of physiological sciences (Diamond, 2004; Noh, 2006; Srinivas, 2010). As obvious from the above description the Bayesian classification method comprises two steps: First, in the learning period, the input of measurement data and known relationships to cardiovascular diseases (training data) are used to train the priors and densities needed in the EM algorithm. The input data can be any type of experimental or computational result that can be related to the model. The probability distributions over the values in these training datasets are learned from examples verified by the gold standard (valid diagnosis), thus allowing the generation of new relationships that describe disease specific classes. The gold standard relationships can include any properties of time series known to be related or unrelated to a particular disease. In general, these gold standards are formed by data obtained for a subpopulation of patients with known diagnosis. Second, in the prediction phase, the classification probabilities are predicted by the classification algorithm based on newly acquired signals (testing data) without available valid diagnosis. The prediction is generally based on a network indicating how likely the observation of measurements fits to a specific class. If one considers this network as a connection matrix, it is just a collection of fuzzy

19


like measures, each representing a probability of functional relationship between the measurement and the class. According to the classification probabilities the procedure provides a diagnostic hint about the existence of afore characterized diseases. In other words, the algorithm sets up a series of hypothesis, that are based on the prior information of a subpopulation obtained in clinical observations, to classify the health condition of the patient by statistical inference. The algorithmic classification procedure is as follows: 1. Use training data d*1:I,1:K to determine optimal parameters λ*, priors and density estimation via the EM algorithm 2. Classify testing data d1†:I ,1:K via (fuzzy) classification probabilities p m (d1†:I ,1:K ) . 3. Integrate testing data into training data set and re-optimize parameters. In order to realize this approach for cardiovascular diseases we will have to train the algorithm on a significantly large population of patients which is the future challenge we will have to face. Then the statistical classification is a method that allows us to identify cardiovascular diseases in an early state that are followed by therapeutic intervention convenient for individual patients. In contrast to other methods, that determine a set of selected parameters with pretended relevance for diagnosis, the classification method automatically selects and quantifies all relevant parameters to prove a series of proposed diseases in the fashion of differential diagnosis.

20

CONCLUSION AND OUTLOOK Within this work we have outlined the problems that may arise in the solution of the constraint hemodynamic inverse problem. We pointed out that ill-posedness is problem inherent and that a unique solution is hindered by (i) the lack of information contained in the finite number of measurements, (ii) the contamination by artifacts and noise and (iii) traditional models and data mining techniques generally used. We have shown that statistical inference methods provide various advantages over deterministic approaches such as quantitative parameter estimates, determination of confidence intervals, treatment of arbitrary forward maps, error estimates and parameter estimates given incomplete and noisy measurement data. These properties are proposed to be the basis in patient specific diagnoses, because they provide a statistical framework for the description of the cardiovascular system that allows therapeutic interventions in individual patients. In contrast to other methods the proposed classification automatically selects and quantifies all relevant parameters to prove a series of proposed diseases in the fashion of differential diagnosis Although the interdisciplinary challenges involved in the ongoing project are daunting, it is important to recognize the potential gains for patient-specific cardiovascular diagnosis. However, up to now the progress has been inhibited by the lack of a broad data basis of non-invasive hemodynamic measurements, advanced inverse modeling tools and databases for large-scale model integration and data classification. Nevertheless we think that the new modeling tools will find several applications in today’s medicine. The progress will depend on the level of support from funding and industry and the interest of clinicians.


REFERENCES Alastruey, J., Parker, K. H., & Peir, Û, J., & Sherwin, S. J. (2008). Lumped parameter models for 1-D blood flow simulations: Effect on pulse waves and parameter estimation. Communications in Computational Physics, 4(2), 317–336. Alastruey, J., Parker, K. H., & Peir, Û, J., & Sherwin, S. J. (2009). Analysing the pattern of the pulse waves in arterial networks: A time domain study. Journal of Engineering Mathematics, 64(4), 331–351. doi:10.1007/s10665-009-9275-1 Avolio, A. P. (1980). Multi-branched model of the human arterial system. Medical & Biological Engineering & Computing, 18(6), 709–718. doi:10.1007/BF02441895 Capova, K., Blazek, V., Cap, I., & Buckuliakova, L. (2002). Physiological fluid systems modelling for non-invasive investigation. Advances in Electrical and Electronic Engineering, 1, 38–42. Ceci, M., Appice, A., Malerba, D., & Colonna, V. (2003). Multi-relational structural Bayesian classifier. AI*IA 2003: Advances in Artificial Intelligence, Lecture Notes in Computer Science. Berlin, Germany: Springer. Chan, I. S., Goldstein, A. A., & Bassingthwaighte, J. B. (1993). SENSOP: A derivative-free solver for non-linear least squares with sensitivity scaling. Annals of Biomedical Engineering, 21, 621–631. doi:10.1007/BF02368642 D’Argenio, D. Z. (2006). Bayesian inference in physiologically-based pharmacokinetic modeling: Application to aniticancer drug development. In D’Argenio, D. Z. (Ed.), Advanced methods of pharmacokinetic and pharmacodynamic systems analysis (Vol. 3). Springer Netherlands.

Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society. Series B. Methodological, 39, 1–38. Diamond, G. A., & Kaul, S. (2004). Prior convictions: Bayesian approaches to the analysis and interpretation of clinical megatrials. Journal of the American College of Cardiology, 43, 1929–1939. doi:10.1016/j.jacc.2004.01.035 Dobrin, P. B. (1978). Mechanical properties of arteries. Physiological Reviews, 58(2), 397–460. Duda, R., Hart, P., & Stork, D. (2001). Pattern classification. John Wiley. Firedman, N., & Kohavi, R. (2002). Data mining tasks and methods: Bayesian classification. In Klosgen, W., & Zytkow, J. (Eds.), Handbook of data mining and knowledge discovery book contents (pp. 282–288). New York, NY: Oxford University Press, Inc. Formaggia, L., Quateroni, A., & Veneziani, A. (2009). Cardiovascular mathematics: Modelling and simulation of the circulatory system. Springer. Gerwinn, S., Macke, J. H., & Bethge, M. (2010). Bayesian inference for generalized linear models for spiking neurons. Frontiers in Computational Neurosciences, 4(12). Hallander, J., Waldmann, P., & Wang, C., & Sillanp, M. J. (2010). Bayesian inference of genetic parameters based on conditional decompositions of multivariate normal distributions. Genetics, 185, 645–654. doi:10.1534/genetics.110.114249 Hartley, H. O. (1958). Maximum likelihood estimation from incomplete data. Biometrics, 14, 174–194. doi:10.2307/2527783

21


Hassani, K., Navidbakhsh, M., & Rostami, M. (2007). Modeling f the aorta artery aneurysms and renal artery stenosis using cardiovascular electronic system. Biomedical Engineering Online, 6(22). Heldt, T., Shim, E. B., Kamm, R. D., & Mark, R. G. (2001). Model-based parameter estimation using cardiovascular response to orthostatic stress. Computers in Cardiology, 28, 337–340. Hoeks, A. P., Brands, P. J., Willigers, J. M., & Reneman, R. S. (1999). Non-invasive measurement of mechanical properties of arteries in health and disease. Proceedings of the Institution of Mechanical Engineers. Part H, Journal of Engineering in Medicine, 213(3), 195–202. doi:10.1243/0954411991534924 John, L. R. (2004). Forward electrical transmission line model of the human arterial system. Medical & Biological Engineering & Computing, 42(3), 312–321. doi:10.1007/BF02344705 John, L. R., & Capper, W. L. (2006). An inverse transmission line model of the lower limb arterial system. Computer Simulations in Biomedicine, 163-170. Jovanovic, M., Colic, M., Stefanovic, P., Ronevic, R., Rasulic, L., & Karapandic, M. (2004). Anatomic analysis of the vascular network and vascular pedicle of the tensor fascia lata flap (angiographic and cadaver study). European Journal of Plastic Surgery, 27, 61–67. doi:10.1007/ s00238-004-0624-z Kaipio, J., & Somersalo, E. (2005). Statistical and computational inverse problems. In Antman, S. S. (Ed.), Applied mathematical sciences (Vol. 160). Springer. Korovkin, N. V., Chechurin, V. L., & Hayakawa, M. (2007). Inverse problems in electrical circuits and electromagnetics. Springer.

22

Lathem, R. D., Westerhof, N., Sipkema, P., Rubal, B. J., Reuderink, P., & Murgo, J. P. (1985). Regional wave travel and reflections along the human aorta: A study with six simultaneous microanometric pressures. Circulation, 72, 1257–1269. doi:10.1161/01.CIR.72.6.1257 Lee, T.-C., Huang, K.-F., Hsiao, M.-L., Tang, S.-T., & Young, S.-T. (2004). Electrical lumped model for arterial vessel beds. Computer Methods and Programs in Biomedicine, 73(3), 209–219. doi:10.1016/S0169-2607(03)00067-1 Manor, D., Beyar, R., Shofti, R., & Sideman, S. (1994). In-vivo study of the mechanical properties of epicardial coronary arteries. Journal of Biomechanical Engineering, 116(1), 131. doi:10.1115/1.2895697 Matthys, K. S., Alastruey, J., & Peir, Û, J., Khir, A. W., Seger, P., Verdonck, P. R., Ö Sherwin, S. J. (2007). Pulse wave propagation in a model human arterial network: Assessment of 1-D numerical simulations against in-vitro measurements. Journal of Biomechanics, 40(15), 3476–3486. doi:10.1016/j.jbiomech.2007.05.027 McCombie, D., & Asada, H. H. (2004). Multichannel blind system identification of the arterial network using a haemodynamic wave propagation model. Proceedings of the American Control Conference, 2, 1645–1646. McCombie, D. B., Reisner, A. T., & Asada, H. H. (2005). Laguerre-model blind system identification: Cardiovascular dynamics estimated from multiple peripheral circulatory signals. IEEE Transactions on Bio-Medical Engineering, 52(11). doi:10.1109/TBME.2005.856260 McLachlan, G. J. (1997). The EM algorithm and extensions. John Wiley.


McNames, J., & Aboy, M. (2008). Statistical modeling of cardiovascular signals and parameter estimation based on extended Kalman filter. IEEE Transactions on Bio-Medical Engineering, 55(1). doi:10.1109/TBME.2007.910648

Olufsen, M. S., & Nadim, A. (2004). On deriving lumped models for blood flow and pressure in the systemic arteries. Mathematical Biosciences and Engineering, 1(1), 61–80. doi:10.3934/ mbe.2004.1.61

Minka, T. (1998). Expectation maximization as lower bound maximization. Retrieved from http:// www-white.media.mit.edu/ tp- minka/papers/ em.html

Pedley, T. J. (2008). The fluid mechanics of large blood vessels. Cambridge University Press.

Mukkamala, R., & Cohen, R. J. (2001). A forward model-based validation of cardiovascular system identification. American Journal of Physiology. Heart and Circulatory Physiology, 281, 2714–2730. Neal, R., & Hinton, G. (1993). A new view of the EM algorithm that justifies incremental and other variants. Technical report, Department of Computer Science. University of Toronto. Neal, R., & Hinton, G. (1998). A view of the EM algorithm that justifies incremental, sparse, and other variants. In Jordan, M. I. (Ed.), Learning in graphical models (pp. 355–368). Kluwer Academic Publishers. Noh, K., Lee, H. G., Shon, H.-S., Lee, B. J., & Ryu, K. H. (2006). Associative classification approach for diagnosing cardiovascular disease. Intelligent Computing in Signal Processing and Pattern Recognition, Lecture Notes in Control and Information Sciences. Berlin, Germany: Springer. Noordergraaf, A., Verdouw, P. D., & Boom, H. B. K. (1963). The use of an analogue computer in a circulation model. Progress in Cardiovascular Diseases, 5(5), 419–439. doi:10.1016/S00330620(63)80009-2 Olufsen, M. S. (2000). A one-dimensional fluid dynamic model of the systemic arteries. Studies in Health Technology and Informatics, 71, 79–97.

Quateroni, A., Ragni, S., & Veneziani, A. (2001). Coupling between lumped and distributed models for blood flow problems. Computational Visual Sciences, 4(2), 111–124. doi:10.1007/ s007910100063 Quick, C. M., Young, W. L., & Noordergraaf, A. (2001). Infinite number of solutions to the hemodynamic inverse problem. American Journal of Physiology. Heart and Circulatory Physiology, 280(4), 1472–1479. Samar, Z., Heldt, T., Verghese, G. C., & Mark, R. G. (2005). Model-based cardiovascular parameter estimation in the intensive care unit. Computers in Cardiology, 32, 635–638. doi:10.1109/ CIC.2005.1588181 Sameni, R., Shamsollahi, M. B., & Jutten, C. (2006). Multi-channel electrocardiogram denoising using a Bayesian filtering framework. Computers in Cardiology, 33, 185–188. Sameni, R., Shamsollahi, M. B., & Jutten, C. (2008). Model-based Bayesain filtering of cardiac contaminants from biomedical recordings. Physiological Measurement, 29(5), 595–613. doi:10.1088/0967-3334/29/5/006 Segers, P., Dubois, F., De Wachter, D., & Verdonck, P. (1998). Role and relevancy of a cardiovascular simulator. Journal of Cardiovascular Engineering, 3, 48–56.

23


Segers, P., Kips, J., Trachet, B., Swillens, A., Vermeersch, S., Mahieu, D., & Van Bortel, Ö, L. (2009). Limitations and pitfalls of non-invasive measurement of arterial pressure wave reflections and pulse wave velocity. Artery Research, 3(2), 79–88. doi:10.1016/j.artres.2009.02.006 Sitz, A., Schwarz, U., Kurths, J., & Voss, H. U. (2002). Estimation of parameters and unobserved components for nonlinear systems from noisy time series. Physical Review E: Statistical, Nonlinear, and Soft Matter Physics, 66, 016210. doi:10.1103/ PhysRevE.66.016210 Smith, B. W., Chase, J. G., Nokes, R. I., Coope, I. D., & Shaw, G. M. (2003). Optimisation based identification of cardiovascular system model parameters for patient specific diagnostic assistance. Sydney, Australia: World Congress on Medical Physics and Biomedical Engineering (WC2003), 24-29 Aug 2003. Srinivas, K., Kavihta Rani, B., & Govrdhan, A. (2010). Applications of data mining techniques in healthcare and prediction of heart attacks. (IJCSE). International Journal on Computer Science and Engineering, 2(2), 250–255. Stergiopulos, N., Westerhof, B. E., & Westerhof, N. (1999). Total arterial inertance as the fourth element of the Windkessel model. The American Journal of Physiology, 276, 81–88. Stergiopulos, N., Young, D. F., & Rogge, T. R. (1992). Computer simulation of arterial flow with applications to arterial and aortic stenosis. Journal of Biomechanics, 25(12), 1477–1488. doi:10.1016/0021-9290(92)90060-E Swamy, G., Ling, Q., Li, T., & Mukkamala, R. (2007). Blind identification of the aortic pressure waveform from multiple peripheral artery pressure waveforms. American Journal of Physiology. Heart and Circulatory Physiology, 292, 2257–2264. doi:10.1152/ajpheart.01159.2006

24

Tardy, Y., Meister, J. J., Perret, F., Brunner, H. R., & Arditi, M. (1991). Non-invasive estimate of the mechanical properties of peripheral arteries from ultrasonic and photoplethysmographic measurements. Clinical Physics and Physiological Measurement, 12, 39. doi:10.1088/01430815/12/1/003 Thore, C.-J., StÂlhand, J., & Karlsson, M. (2008). Toward a noninvasive subject-specific estimation of abdominal aortic pressure. American Journal of Physiology. Heart and Circulatory Physiology, 295, 1156–1164. doi:10.1152/ajpheart.01154.2007 Timmer, J., Rust, H., Horbelt, W., & Voss, H. U. (2000). Parametric, nonparametric and parametric modeling of a chaotic circuit time series. Physics Letters. [Part A], 274, 123–134. doi:10.1016/ S0375-9601(00)00548-X Tong, L., & Perreau, S. (1998). Multichannel blind identification: From subspace to maximum likelihood methods. Proceedings of the IEEE, 86(10). Wang, J.-Z., Tie, B., Welkowitz, W., Kostis, J., & Semmlow, J. (1989). Incremental network analogue model of the coronary artery. Medical & Biological Engineering & Computing, 27(4), 416–422. doi:10.1007/BF02441434 Wessel, N., Voss, A., Malberg, H., Ziehmann, Ch., Voss, H. U., & Schirdewan, A. Kurths, J. (2000). Nonlinear analysis of complex phenomena in cardiological data. Herzschrittmachertherapie & Elektrophysiologie, 11(3), 159–173. doi:10.1007/ s003990070035 Westerhof, N., Bosman, F., De Vries, C. J., & Noordergraaf, A. (1969). Analog studies of the human systemic arterial tree. Journal of Biomechanics, 2(2), 121–143. doi:10.1016/00219290(69)90024-4


Womersley, J. R. (1955). Oscillatory motion of a viscous liquid in a thin-walled elastic tube. I. The linear approximation for long waves. Philosophical Magazine, 46, 199–221. Xiao, X., Ozawa, E. T., Huang, Y., & Kamm, R. D. (2002). Model-based assessment of cardiovascular health from noninvasive measurements. Annals of Biomedical Engineering, 30, 612–623. doi:10.1114/1.1484217 Zamir, M. (2000). The physics of pulsatile flow. Springer Biological Physics Series. Zanchi, A., Brunner, H. R., & Hayoz, D. (1997). Age-related changes of the mechanical properties of the carotid artery in spontaneously hypertensive rats. Journal of Hypertension, 15(12), 1415–1422. doi:10.1097/00004872-199715120-00008 Zhang, Y., & Harry, A. H. (2004). Blind system identification of noncoprime multichannel systems and its application to noninvasive cardiovascular monitoring. Journal of Dynamic Systems, Measurement, and Control, 126(4), 834–847. doi:10.1115/1.1852460

KEY TERMS AND DEFINITIONS Hemodynamic Forward Problem: Within the forward problem the pressure and flow waveforms at places of interest are computed for the applied input pressure or flow using a set of model parameters for the circulatory system. This implies that the lengths, radii and compliance values of all vessel segments are known.

Hemodynamic Inverse Problem: Within the inverse or backward problem the model parameters of the circulatory system are estimated when pressure and flow at specific places are given. Well-posed - Ill-posed: The mathematical term well-posed defines that a unique solution to a mathematical model of a physical phenomena exists, that depends continuously on the data. Problems that are not well-posed are termed illposed. Due to the lack of information in measured data inverse problems are often ill-posed. Condition: Even if a problem is well-posed, it may be ill-conditioned, meaning that the results are strongly dependent (sensitive) on the initial data. An ill-conditioned problem is indicated by a large condition number. Regularization: If the problem is not wellposed, it is generally re-formulated for numerical treatment. Typically this involves including additional assumptions, such as smoothness of solution. This process is known as regularization. Bayesian Inference: is statistical inference in which observations are used to infer the probability that a hypothesis may be true. Stenosis and Aneurism: A stenosis is a constriction of blood vessels mostly based on arteriosclerosis, while an aneurism is the term for a balloon-like dilation of a vessel. Both are vascular defects, the former can cause ischemia, while in the latter case the vessel may burst under normal pressure conditions. Compliance: The term that quantifies the dispensability of a blood vessel for a specified pressure increment.

25

26

Chapter 2

Personalized Experience Sharing of Cai’s TCM Gynecology Guo-Zheng Li Tongji University, China Mingyu You Tongji University, China Liaoyu Xu Shanghai Literature Institute of TCM, China Suying Huang Shanghai Literature Institute of TCM, China

ABSTRACT This chapter introduces experience sharing of Cai’s gynecology in Traditional Chinese Medicine (TCM). It argues that Cai’s family in China are experts in TCM gynecology, whose seventh generation Xiaosun Cai is a TCM master. Therefore, inheriting experience of Xiaosun is a critical business in the present TCM field. The authors demonstrate a novel system (named TCM-PMES) for preserving the diagnostic processes of veteran practitioners like Xiaosun in a personalized way. Based on the summarized Cai’s diagnostic template, a custom input system embracing the particular expressions of the specialist is set up in the platform. Unique and unfiltered experience of veteran practitioner is maintained as complete as possible. Various approaches on exploring relationships among TCM components and speculating syndrome types from clinical symptoms are also studied. This work provides a new way for keeping and researching the personalized factors in the diagnostic process of Cai’s TCM gynecology.

DOI: 10.4018/978-1-61350-120-7.ch002


Personalized Experience Sharing of Cai’s TCM Gynecology

INTRODUCTION Gynecology is so complicated, that doctors often say that illness of women is ten times harder to cure than that of men (Cai, Huang & Mo, 2000). Different from men, illness of women is related with menstruation, pregnancy, vaginal discharge and labor. Due to the variance of women with different conditions, the pathological mechanism and treatment prescription are not the same in TCM (Traditional Chinese Medicine) gynecology. Even for cold headache or pharyngalgia, if the patient is in the period of menstruation, we need to consider whether she is in disorder of menstruation, the quantity or abdominalgia. The prescription should not only release the exterior or disturb the menstruation, but also accord to the therapeutic principle of considering the root and tip in total. If the same disease takes place in another pregnant woman, some herbs hurting the foetus must not be used in principle, but may be used in specific cases. If the ill women newly gave birth, they have much spontaneous perspiration. The doctor should consider they are blood deficiency and need to discharge the lochia completely. So relieving the exterior with pungent and pool need to be improved with other measures. In TCM gynecology, we further need to consider their physical type, old or young, living conditions, working status and so on of the patients. Cai’s gynecology started from 250 years ago, the present existing successor is the 7th, whose name is Xiaosun Cai. Xiaosun was born in 1923, he has dedicated to medical care for more than 60 years. Xiaosun was elected as the national tutor for inheriting the maters’ experience in 1992 and included in the medical who’s who published by Cambridge Press in 1995. Now he has several students, one of which is Suying Huang, one of the authors. Xiaosun has inherited and accumulated ample experience in his career, he still takes front line jobs now. TCM masters like Xiaosun are high years old, and their experience may be lost some day, which are inherited from their family or ac-

cumulated in their work throughout their lives. These experiences are valuable, and hard to be summarized into knowledge directly now due to the complex TCM principles. Therefore, collecting their medical records and mining their experience is a critical work in the TCM field, which is also a policy of the Chinese government. As a complete medical system, TCM plays an indispensable role in medical care in China for more than five millenniums. It is a treasure of Chinese culture. Different from the Reductionism thinking mode of Western Medicine (WM), TCM is based on the holistic and systematic ideas. Classification of TCM syndrome should be based on overall analysis of symptoms and signs of human body. Symptoms or signs in different areas of body may contribute to the same kind of syndrome. Even the relationship between disease and syndrome might be many-to-one. Based on the methodology of holism, TCM plays a unique role in health care and disease treatment. Increasing usage of Chinese natural herbal medicine and acupuncture world widely is a good indication of effectiveness of TCM (Yamashita, Tsukayama, & Sugishita, 2002; Honda & Jacobson, 2005; Thomas, Nicholl, & Coleman, 2001). Medical knowledge and understanding of TCM veteran practitioner is the practical and creative achievement of TCM principles. They are all valuable assets to the TCM system. Knowledge of veteran practitioner on TCM clinical diagnosis and Chinese herbal medicine sometimes cannot be found in literatures. Traditionally, the precious medical knowledge and philosophical view of veteran practitioner is followed by his apprentices (C. Zhou, 2004). The number of juniors who have the honor to learn from a great master is limited. Furthermore, whether the juniors can grasp the soul of the knowledge system of the veteran practitioner totally depends on his personal ability, sometimes even gift. As most of the famous TCM practitioners getting older, preserving and researching into their clinical records, books or publications becomes urgent and important.

27


Preserving the knowledge of the TCM veteran practitioner includes understanding, systematical sorting and standardized inputting the available materials from the specialist. A powerful platform is the basis for literature and data gathering. Many TCM online systems are reported in published works. TCMMDB (X.-Z. Zhou, Wu, & Lu, 2001) is a unified web accessible multi-database query system, which has already integrated more than 50 databases. TCM-Grid (Chen, Wu, Huang, & Xu, 2003) is a grid-based TCM system. It is also extended to a semantic-based database grid. X. Zhou et al. (2004) have developed a large-scale ontology-based Unified TCM Language System to support concept-based information retrieval and information integration. These three systems mainly contribute to the collection of Chinese drug database or Chinese medical literature. They pay little attention to the clinical records of TCM practitioners, which are practical experience of TCM principles. Zhang et al. (2008) designs a platform for analyzing and mining knowledge of TCM veteran practitioner. The system is set up with a business software BO. All the platforms or systems mentioned above provide uniform input items for different veteran practitioners. Different from strict standardization of WM, clinical records from TCM practitioners, especially veteran practitioners, have obvious and splendid personal focus and expression. The symptoms and signs one practitioner cares might be different from the other in most parts. The herbal medicines and their combination that veteran practitioner prefers also have distinctly personal factors. All the personal items in the clinical process may be the embodiment of the TCM veteran practitioner’s thinking mode and knowledge system. To preserve the symptoms, signs, formula and so on in their original taste and flavor is an interesting and significant task. Upon the numerous TCM databases, different researches are conducted to discover the implied knowledge. Chinese Medical Formula (CMF) is composed of compound drugs with suitable doses.

28

The potency of a single Chinese herb is limited. Application of several natural herbs ensures a full play of their advantages and inhibiting some side effects. Among the combinations, when two kinds of herbs are frequently used together, they are likely to be paired herbs. The number in paired herbs may be extended to three, four, etc. Yao, Ai, Yuan, and Qiao (2002) use the classical association rule method to study 106 formulae for treating diabetes. The results indicate that different experts have similar ideas and principles for this disease. Zhu et al. (2007) also adopt association rule to analyze the medical records on asthma. The incidence relation among pathogeny, location of disease and syndrome with four kinds of symptoms, the relationship among pathogeny, location of disease, syndrome, four kinds of symptoms and the use of TCM, the relationship among the TCM, are all summarized. Besides association rules, different methods and models are provided to discover the knowledge for TCM clinical diagnosis. Qin, Mao, and Deng (2007) use rough set in the diagnosis of rheumatoid arthritis. The results show that the diagnostic accuracy achieved by rough set for rheumatoid arthritis is greatly higher than that of fuzzy set. A novel self-learning expert system for diagnosis in TCM is constructed by using a hybrid system composing of a Bayesian network learning algorithm, naïve-bayes classifiers with a novel score-based strategy for feature selection and a method for mining constrained association rules. The learned knowledge is provided in multiple forms including causal diagram, association rule and reasoning rules derived from classifiers (Wang, Qu, Liu, & Cheng, 2004). Supported by the Shanghai government, we launched a project to collect 10 TCM masters’ medical records, where these masters are all over 75 years old. We designed a personalized TCM platform for Maters’ Experience Sharing (named TCM-PMES) (You, Ge, Li, Xu, & Huang, 2009; You et al., 2008). This chapter aims to provide TCM-PEMS to preserve the precious materials of Cai’ gynecology. We further introduce the sum-


mery of Cai’s experience. A brief introduction of the main functions and structure of TCM-PMES is presented in Section 2. Clinical records from Xiaosun Cai are included to illustrate the process of summarizing the clinical template and setting up the personalized system in Section 3. In Section 4, several interfaces of the set up system are explained. In Section 5, experience summarized from the above medical records are presented and discussed. Finally, the chapter concludes on the future work in Section 6.

TCM-PEMS: A Personalized TCM Platform for Masters’ Experience Sharing Cai’s gynecology came from 250 years ago, it has many unique terms to describe the disease, the symptoms, cause of disease, mechanism of disease, diagnostic methods, prescription or medical formula, which are different the general terms normalized by the government or international societies. We design a personalized TCM Platform for Maters’ Experience Sharing (TCM-PEMS), whose functional diagram is shown in Figure 1. The whole system is based on B/S architecture. Users visit the client by web browser. The server is composed of logic layer and database. The database stores specialist, herbs, syndrome, symptom, clinical record and CMF (Chinese Medical Formula) information. The discovered knowledge is also stored in the database. The client supports five kinds of user roles, and different role user has different authority. Standard user retrieves clinical record and CMF by given arguments, such as disease or symptom. Standard user also predicts the most possible syndrome and CMF from the symptom by leveraging the discovered knowledge saved in database. The knowledge models summarize the occurrence frequencies of symptoms, syndromes and herbs. The knowledge models also abstract the relationship between symptoms and syndromes, the relationship between syndrome

and herbs, the relationship between special symptoms and combination of basic formulas, and the internal relationship of herbs. Data mining methods (Li, 2010) have been already implemented in PTCMS include association rule analysis, support vector machine and so on. Knowledge models are trained by data mining operators, who have the authority of mining clinical records and CMF. The training data is input and preprocessed by data entry assistant. The data maintenance assistant maintains the database. Administrator manages users and monitoring the whole system. The platform is composed of five main layers in function (Figure 2). Data layer contains databases storing information of medicine, CMF, symptom, syndrome, clinical record and specialist. Knowledge models and learning algorithms are saved in the knowledge DB in data layer. The clinical record DB saves personalization template and clinical data. The syndrome DB stores the records of disease, syndrome and therapies with TCM. The specialist DB records the specialist information of individual, learning experience, academic background, academic thinking and clinical speculation. Symptom DB, CMF DB and medicine DB include essential information of symptom, CMF and medicine, respectively. In persistence layer, hibernate is an ORM (objectrelational mapping) library for the Java language, which solves object-relational impedance mismatch problems by replacing direct persistencerelated database accesses with high-layer object handling functions. The top three layers comprising business, action and view, are managed with the Model-View-Controller (MVC) pattern. The business layer manages the behavior and data in the application domain. There are three parts of application domains in the platform. Firstly, in the domain of statistic and analysis for data, the system accounts the frequencies of occurrence of symptoms, syndromes and herbs, analyzes the prescribing habits of specialists. Besides, the system abstracts the relationships among symp-

29


Figure 1. Functional diagram of TCM-PMES

toms, syndromes, herbs and CMF by knowledge models. Secondly, in the clinical management domain, the information of symptom, syndrome, medicine and CMF is maintained. Clinical records including symptom, syndrome, and CMF can be input into the clinical page, which is generated according to the personalized template. Besides data recording, the knowledge models can predict the most possible syndrome from the symptoms of a patient and suggest a suitable CMF. Finally, the content of the specialist domain is a range of specialist information, such as learning experience, academic background, academic thinking

30

and clinical speculation. The action layer interprets the mouse and keyboard inputs from the user, informing the business layer and the view layer to react appropriately. The view layer manages the display of different information.

Diagnostic Template for Cai’s Gynecology Due to the unique terms in Cai’s Gynecology, we design a personalized diagnostic template. Cai’s Gynecology is famous in China, even in the world, whose representative is Mr. Xiaosun


Figure 2. System architecture of TCM-PMES

Cai in present. Xiaosun is now 87 years old, he is the seventh generation of “Cai’s Gynecology”. The characteristics of Cai’s Gynecology is that the four diagnostic methods: inspection, listening and smelling, inquiry, and palpation are all very important, which need a comprehensive consideration of all four methods (Cai,2000). “Nine Inspection” is summarized as inspecting the face color, tongue color, lip color, nail, menstrual color and quality, leucorrhea, lochia, hair, as well as others like philtrum. For facial color, if one is blood deficiency, she will be chlorosis;

if one is qi deficiency, she will be pallor or superficial; if one is blood stasis, she will be dark blue. If her face is red in cheek, she is yin deficiency. Tongue color and fur are also important, if tongue color is fresh red, she is blood heat; else if color is light, she is blood deficiency; else if color is yellow and fur is slimy in its root, she has dampness and heat in her lower energizer; else if there are purple points in margin of tongue, then she is blood stasis. Lip color reflects the status of spleen, if it is red and purple, she is blood heat; else if it is fresh red and crack, she is yin deficiency

31


with effulgent fire; else if it is light, she is spleen deficiency and blood deficiency; else if it is deep blue, she has pain, while slight blue means she is deficiency and cold. Menstruation is a critical factor, if its color is fresh red or deep red, she is blood heat; else if its color is slight red, she is deficiency; else if its color is dark purple; she is blood stasis. If its property is thick, she has heat; else if it is thin, she is deficiency with cold; else if it is with blocks of blood, she is blood stasis. Virginal discharge reflects the status of one’s reproductive system. If it is white and thin property, she is spleen deficiency; else if it is yellow with thick property, she is dampness and heat in lower energizer; else if it is blue with thin property, she is kidney deficiency; else if different colors appear, she has tumor in large possibility. In TCM, philtrum is an important position to reflect the status of reproductive system. If it is flat, her womb may not well cultured; else if it is narrow, her womb cervix is also narrow; else if it is short, her womb cervix may be short; else if there is acne on it, there is blood stasis in her reproductive system. “Listening and Smelling” means ausculating the voices and other sounds and sniffing the tastes. If her voice is low and thin, she is qi deficiency; else if she often sigh, she is qi depression; else if her menstruation is foul smelling, she is blood heat; else if her virginal discharge is foul smelling, she is dampness and heat in lower energizer. “Eight Inquiring” includes inquiring the age/ marriage status, main symptoms, medical history, menstrual cycle, leucorrhea, pregnancy reaction, post partum, and professional life. The age and marriage status are important information. If there is irregular bleeding, she is functional womb bleeding for a girl, while she is malign tumor for a old woman. Some women is shame to say her symptoms or her status, e.g. sex life. Unmarried girl says she is delayed menstruation. But if she is ectopic pregnant, it is danger. As we can see in the above, we need to know the quantity, color and property of menstruation and virginal

32

discharge. We also need to know her professional life, whether she has over nervous work or study. We need to know her diet and spirits, etc. “Nine palpation” comprises palpating the menstrual pulse, leucorrhea pulse, pregnancy pulse, parturient pulse, postpartum pulse, skin, head and limb, breast, abdomen. If one is during the menstruation, her pulse should be more slippery and surging than usual. If the pulse is hollow, it means her qi and blood drop suddenly, which occurs when flooding or giving birth. For the pulse of a woman with virginal discharge, if it is spring like and racing, she is dampness and heat; else if it is sunken and thin, she is kidney deficiency; else if it is spring like and slippery, she has tumor or inflammation. We should touch the skin of patients, if it is not warm in four limbs, she is qi and yang deficiency; else if it is hot in hands and foots, she is yang and heat excess; else if it is hot in the center of hands and foots, she is yin deficiency and heat interior. On the abdomen, we need to know whether the napes is soft or hard, warm or cool, whether there is blocks and their location, size and property. If the block is hard and painful, she is blood stasis; else if there is no block or she is painless, she is blood deficiency; else if the block disappears when be pushed, she is qi stasis. The above four diagnostic methods are principles of ”Cai’s Gynecology” in hundreds of years. Now they claim to acquire an overall and deep knowledge of the disease from macro to micro, combining the modern scientific instruments with the four techniques. For the gynecology examination, the vulva is important to know the quality, color, smell of the secretion and others. The inner reproductive system is also needed examination by considering both the vagina and abdomen. This study summarizes a basic template from 320 clinical cases of Xiaosun. It is demonstrated from Table 1 to Table 6. We separate the information of patients into five parts, i.e. basic part, cardinal part, general menstrual part, local menstrual part and WM normal examination part. There are


Table 1. Basic information part in the personalized template of Cai’s Gynecology Patient Name Gender

Male/Female

Age Medical Record No. First Visit Date Menstrual History

Age of Menarche; Menstrual Period; Menstrual Duration

Marriage Status

Single/Married (YOM year of marriage)

Child Birth History

Times of pregnancy; Times of Miscarriage; Times of Abortion; Number of Children

Contraception

Yes (Condom/Oral Contraceptive Pills/Intrauterine Device)/No (How Long)

LMP (last Menstruation Period) PMP (Preparatory Menstruation Period) BBT (Basal Body Temperature)

Single Phase/ Two Phase/ Two Phase but Nontypical/ Rising/Duration of High Temperature Phase Less Than 12 Days/ Can Not Rise Stably/ BBT Rise and Drop/ BBT Rise but Do Not Drop

Chief Complaint History of Present Illness

Table 2. Cardinal symptoms part in the personalized template of Cai’s Gynecology Status of Patient

Pre and Post the Menstrual Phase/ Latecycle/ Midcycle (Near/ Exact/ Late); Near the Menstrual Phase; Late Periods; Period Stops for More Than 2 Months; Amenorrhea; Distended Pain (Acid Swells) in Right/Left Lower Quadrant; Sharp Pain in Lower Quadrant; Aching Pain in the Two Sides of Lower Quadrant

Menstrual Period

Regular Periods; Irregular Periods Recently; Early Periods; Late Periods; Irregular Periods; Amenorrhea; Regular Period Last time; Just Begin the Menstrual Phase; Near the Menstrual Phase; Over Due

Amount of Menstrual Bleeding

Pour; Collapse; Heavy; Medium; Fluent; Unimpeded; Impeded; Scanty; Drop Off; One Day Off; Few; Increase; Drip Several Days

Color of Menstrual Bleeding

Slight Dark at the Beginning; Bright Red; Dark Red; Black; Coffee Color

Thickness of Menstrual Bleeding

Thick; Thin; Lump; Filmy; Lump and Filmy

Odor of Menstrual Bleeding

Yes/No

Painful in Periods

Severe Pain in the Lower Abdomen; Distended Pain; Slight Pain; Dull Pain; Sharp Pain; Pain in the Lower Back; No Pain in the Abdomen

Methods for Pain Relief

Relief after Silt Disappear; Relief whenWarming; Relief when Pressing

Leucorrhea

Amount Large/ Small; White/ Yellow/ Green/ Yellowgreeen/ Red/Slight Red/ Coffee Color/ Purple Color/ Slight Purple; Fiber Drawing

Table 3. General accompanying symptoms part of the personalized template of Cai’s Gynecology Exhaustion; Debilitation; Aversion to Cold; Thermophilic; Sweat when Week Up in the Morning; Dysthesia; Fever Disappear after Menses and Hot Flush; Feeling Cold before Menses; Dry Mouth; Drowsy; Baking Fever with Sweat; Baking Fever at Night; Endogenous Heat; Bit Inspiring; Aversion to Wind; Sleeping Badly; Dizziness with Pitching; Face without Floursh; Sallow Complexion; Distended Pain in the Vertex; Weight in Rush; Becoming Fat; Normal Sleep

33


Table 4. Local accompanying symptoms part in the personalized template of Cai’s Gynecology Head

Headache before Menses; Dizzy; Blurred Version; Migraine in the Menstrual Period; Scrofulosis in Face; Alopecia; Dense Hair; Pain in Forehead Sometimes; Blush; Oral Ulceration; Fetid Breath; Throat Staying with Sputum

Chest

Chest Stuffiness; Engorgement of Breast before the Menses; Sharp Pain in Breast; Feeling Tenderness Like Crawling under Breast; Travail in the Heart Area Recently

Stomach

Nausea; Emesis; Distended Abdomen; Tumult in Stomach

Waist

Lumbago; Soreness of Waist (Slight/ Bitter/Broken); Discomfort in Lumbar Vertebra; Ache and Dropping in Waist, Heavy Waist, Waist Suffering which Prefers Heating

Abdomen

Psychroalgia; Dull Pain; Attacked when Being Cold; Pain Sometimes; Abdominal Distention in Right/Left Side; Dull Pain in Both Sides of Abdomen; Distending Pain around Waist and Double Sides of Hypogastrium; Dull Pain in the Lower Abdomen; Sensation of Fullness in Lower Abdomen; Dyspareunia; Burning Pain in Both Sides of Ovarian Area; Tough Pain in Abdominal Wall; Slight Pain in Liver Area; Slight Dropping in Abdomen

Four limbs

Aching Pain in Arm; Thirsty and Loss of Appetite; Limb Feeling Light; Aching Pain in Knee; Calcination in the Foot Center at Night; Calcination in the Hand and Foot Center ; Limb Feeling Soft; Sour in Legs; Sour and Cold in Legs; Talalgia when Laboring

Genitalia and Anus

Nonsolid Stool; NonFluent Stool; Constipation; Stool Every Two Days; Stool Everyday; Diarrhea; Solid Stool; Easy Relieve Oneself; Anus Bearing-down; Dyschizia with Flart; Pollakisurie; Pruritus Vulvae; Dysuria; Micturition Desire Immediately after Drinking; Urinary

Table 5. Examination part in the personalized template of Cai’s Gynecology HCG

Positive;Weak Positive; Negative

B Ultrasonic Hysterosalpingography Examination of Endocrine Function

Table 6. Diagnosis part in the personalized template of Cai’s Gynecology Diagnosis of TCM Dysmenorrheal (Primary/ Secondary); Amenorrhea; Infertility (Primary/ Secondary/ All the time/ Some time); Irregular Periods; Metrorrhagia; Scanty Periods; Abortion; Miscarriage; GynecologicalMass in Abdomen; Postoperation of GynecologicalMass; Migraine in the Menstrual Period; Metrostaxis; Symptoms During Menopause; Stranguria; Fluor; Leukorrhea with Bloody Discharge Diagnosis of WM Endometriosis (Adenomyosis/ Chocolate Cyst); Hysteromyoma; Primary Dysmenorrheal; Secondary Dysmenorrheal; Amenorrhea; Abortion; Postoperation of Chocolate Cyst Decollement; Threatened Abortion; Irregular Periods; Ovarian Cyst; Dysfunctional Uterine Bleeding; Climacteric Syndrome; Menopausal Gong Blood; Chronic Pelvic Inflammatory Disease; Pelvic Congestion Syndrome; Colpitis Mycotica

14 items in the basic part, i.e. name, gender, age, record number, first visit data, menstrual history, marriage status, child birth history, contraception, last menstruation period, preparatory menstruation period, basal body temperature, chief complaint

34

and history of present illness, which are summarized in Table 1. Cardinal symptoms are those main factors which make major effect when the doctor decides the syndrome, therapeutic principle and method.


Table 7. Syndrome part in the personalized template of Cai’s Gynecology 1.ChongRenShiTiao; 2.ShenQiBuZu, LuoDaoBuTong; 3.ShenQiBuZu (ShenQiKuiSun); 4.GanShenBuZu, ChongRenShiTiao; 5.XinShenBuzu; 6.SuYuNeiJie, GanShenBuZu; 7.SuYuNeiJie, LuoDaoQianChang (YuZuLuoDao); 8.YuReNeiYun; 9.SuYuNeiJie, ShenQiBuZu; 10.ShenQiBuZu, LuoDaoQianChang, ShiReXiaZhu; 11.WeiQiBuGu, QieYouGanFeng; 12.SuYuNeiJie, JiErChengZheng (YuKuaiNeiJie); 13.YingYinBuZu, ChongRenQianTiao; 14.FeiShiXuanJiang; 15.ShenQiBuZu, ZhiMoYongZhi (ShenQiBuZu, TanZhiYongZhi); 16.GanYuPiXuShiShe; 17.ShenQiBuZu, ShiYuLuoDaoQianChang; 18.ShenXuXueYu, ShiReXiaZhu; 19.TiXuWeiFu, SuYouYuJie; 20.ShenXuChongRenShiTiao; 21.PiShenBuZu, TanXuYongZhi, ChongRenShiTiao; 22.GanShenYuRe, ChongRenShiTiao; 23.PiShenQiXuShiShe; 24.PiXuShiShe; 25.GanShenBuZu, XueBuYangGan, ShenXuBuJu; 26.GanHuoPianWang; 27.GanYuShenXu; 28.GanWangPiXu; 29.GanPiBuTiao; 30.HanNingBaoGong; 31.XinShenBuZuJianGanYu; 32.GanYuPiXu, XuReZhuoLuo; 33.ShiReXiaZhu; 34.ShiReXiaZhu, RiJiuShangYin; 35.YingXueBuZu, QiXuBuShe; 36.XueXuGanWang, ShiReXiaZhu; 37.YinXuXueRe, ChongRenShiTiao; 38.XueXuShenKui, TanZuBaoLuo; 39.QiXuJiaYu, ChongRenBuGu; 40.QiXuBuZu, ChongRenLiangKui; 41.QiXueLiangWang, ChongRenShiTiao

Table 8. Therapeutic principle part in the personalized template of Cai’s Gynecology 1.TiaoLiChongRen; 2.YuShenTongLuo; 3.YuShenPeiYuan; 4.YuShenTiaoLi; 5.YangXinYuShen; 6.YuShenTiaoLi; 7.HuaYuTongLuo (QingYuTongLuo); 8.QingReTiaoJing; 9.FuZhengQingYu (FuZhengQuYu); 10.YuShenTongLuo, QingReLiShi; 11.FuZhengPingGanQuFeng; 12.HuaYuSanJie (JieYuXiaoJian, QingYuTiaoLi); 13.ZiYinTiaoChongRen; 14.XuanJiangFeiQi; 15.YuShenHuaZhi, LiQiTongLuo; 16.JianPiShuGanTiaoShe; 17.YuShenTongLuoTiaoJing; 18.TiaoLiChongRen, ZuoXieHuoZhiDai; 19.FuZhengGongGu; 20.JianShenTiaoJing (YuShenTiaoJing); 21.LiQiHuaZhi, TiaoChongRen; 22.ShuGanXieHuo, TiaoLiChongRen; 23.TiaoShePiShen; 24.JianPiYiQiGuShen (YiQiTiaoShe); 25.JianLiGanShen; 26.ShuGanXieHuo; 27.YuShenShuGanTiaoLi; 28.NingShenShuGan, PingGanHuanJi; 29.JianGuGanPi; 30.WenGongTiaoJing; 31.BuShenYangXinJianShuGan; 32.JianPiShuGan, YangYinQingHuo; 33.QingReLiShiZhiDai; 34.QingReLiShi, YiQiYangYin; 35.BuYiYangXueTiaoShe; 36.LiShiXieHuo; 37.QingYingTiaoGu; 38.YanXuePeiYuan, QuYuTongLuo, HuoXueTiaoJing; 39.YiQiTiaoGu, QuYuShengXin; 40.BuQiYangXue, JianGuPiShen; 41.YiQiYangXueTiaoGu

There are 9 items, i.e. patient status, menstrual period, amount of menstrual bleeding, color of menstrual bleeding, thickness of menstrual bleeding, odor of menstrual bleeding, painful in periods, pain relief methods and leucorrhea. The detail description and properties of the above 9 items are listed in Table 2. All the symptoms are direct symbols of gynecological disorder and can be observed by patients themselves. Besides the basic information and cardinal symptoms, there are accompanying symptoms, which are important but not critical. As in Table 3, we summarize 22 items like exhaustion, debilitation, aversion to cold, thermophilic, sweat when week up in the morning, dysphasia, fever disappear after menses and hot flush, feeling cold before menses, dry mouth, drowsy, baking fever with sweat, baking fever at night, endogenous heat, bit inspiring, aversion to wind, sleeping badly, dizziness with pitching, face without flourish, sallow complexion, distended pain in the vertex, weight in rush, becoming fat and normal sleep as the general accompanying symptoms,

since these occurs in the body. While the accompanying symptoms in the local part of our body are summarized as local accompanying part like head, chest, stomach, waist, abdomen, four limbs, genitalia and anus, pulse and tongue. The above 9 large items have many sub items which are described in detail in Table 4. Since WM is in the major position in China, many patients conduct physical examinations before they visit the TCM physicians. The examination indices may give some information of the patients, so Xiaosun accepts these indices as some symptoms. The commonly used indices are HCG, B ultrasonic, hysterosalpingography and endocrine function, which are summarized in Table 5. Disease names of TCM and WM used in ”Cai’s Gynecology” are presented in Table 6, where there are 15 TCM names and 15 WM names. There are major difference between the expression of diseases in TCM and WM. Table 7 lists 41 TCM syndromes frequently appearing in the clinical records of Xiaosun. At the

35


Table 9. Basic formula part in the personalized template of Cai’s Gynecology YuShenPeiYuanFang

FuLing (YunLing, YunFuLing) 12g; ShuDi 10g; ShengDi (DaShengDi) 10g; XianMo 10g; XianLingPi 10g; BaJiTian 10g; RouCongRong 10g

SiWuTiaoChongTang

ChaoDangGui 10g; ChuanXiong 6g; ShengDi 10g; BaiShao 10g; ZhiXiangFu 10g; HuaiNiuXi 10g

YuShenTongLuoFang

YunLing 12g; ShengDi 10g; ShuDi 10g; LuLuTong 10g; JiangXiangPian 3g; XianLingPi 12g; ZhiHuangJing 10g

NeiYiYiFang

ChaoDangGui 10g; DanShen 10g; ChuanNiuXi 10g; ZhiXiangFu 10g; ChuanXiong 10g; ChiShao 10g; ZhiMoYao 6g; YanHuSuo 12g; ShengPuHuang (BaoJian) 12g; WuLingZhi 10g; XueJie 3g

NeiYiErFang

DangGui 10g; ShengDi 10g; DanShen 10g; BaiShao 10g; XiangFu 10g; ShengPuHuang 30g; HuaRuiShi 20g; ShuJunTan 10g; SanQiMo (Tun) 2g

NeiYiSanFang

YunFuLing 12g; GuiZhi 3g; ChiShao 10g; DanPi 10g; TaoRen 10g; ZaoJiaoCi 30g; ZhiJiaPian 9g; ShiJianChuan 20g; EShu 10g; ShuiZhi 6g

HuaYuXiaoJianFang

YunLing 12g; GuiZhi 3g; ChiShao 10g; DanPi 10g; EShu 10g; ShanJiaPian 10g; ZaoJiaoCi 30g; PuGongYing 14g; ShiJianChuan 20g; HaiZao 12g; ShuiZhi 6g; QingPi 5g; ChenPi 5g

same time, 41 therapeutic principles are proposed, each of which is to treat the TCM syndrome with the same serial number as in Table 8. Here the TCM terms used by Xiaosun are unique and most of them are not used by other physicians. So the translation job is hard, we have to use Pinyin in Chinese to express them. The therapeutic principles are combinations of basic TCM therapeutic ones, while most of them have one CMF listed below. We will further introduce them by using one case of womb tumor later. Seven basic TCM formulas are summarized in Table 9. One of the basic formulas is selected for a patient based on her TCM syndrome diagnosis. The selected basic formula is then modified according to the patient’s individual symptoms. Modifications include adding/deleting one of the herbs, or increasing/decreasing the doses.

Implementation of TCM-PEMS Using the personalized template of diagnosis process of Cai’s Gynecology, a unique system for preserving the clinical records is set up. Detailed illustration is presented in this section.

36

Process of Setting up a Personalized System With TCM-PEMS, we set up a personalized system for Cai’s Gynecology as the following steps. •

• •

Step 1. Select the entry Inquiry and Diagnosis Template Maintenance in the module tree to go to its main page as in Figure 3. Click button Add to pop up the New Template dialog. Step 2. Input the template number and name. Step 3. Input the standard or nonstandard TCM diseases and syndromes. ◦âŠ¦ Step 3.1. Standard TCM diseases Check the option Use standard TCM disease. Select a disease from the TCM diseases database as the default value. If there is no corresponding record for certain standard disease and syndrome in the database or the information isn’t correct, you can add or modify them in the TCM Disease Maintenance and TCM syndrome Maintenance modules 3. ◦âŠ¦ Step 3.2. Nonstandard TCM diseases Uncheck the option Use standard


Figure 3. Choose to use standard names of TCM diseases

Figure 4. Input a nonstandard TCM disease

Figure 5. Edit the nonstandard TCM diseases

TCM disease. Input the TCM diseases and syndromes manually. As several diseases may have the same syndromes, users can fill in multiple diseases separated by comma as in Figure 4. To help users input the correct format, an input dialog will pop up if they click the disease edit box as in Figure 5. Click Add to insert a disease. Click Modify to modify a disease. Click Delete to delete a disease. Similarly, an input dialog will pop up if they click the syndromes edit box as in Figure 6.

•

Step 4. Input the therapeutic principles. ◦âŠ¦ Step 4.1. Standard therapeutic principles Check the option Use standard therapeutic principle. Select a record from the therapeutic methods database as the default value. If there is no corresponding record for certain standard method in the database or the information isn’t correct, you can add or modify it in the Therapeutic Principle Maintenance module. ◦âŠ¦ Step 4.2. Nonstandard therapeutic methods Uncheck the option Use standard therapeutic principle. Input 37


Figure 6. Edit the syndromes of a nonstandard TCM disease

Figure 7. Edit personalized therapeutic principles

38


Figure 8. Select the type of WM diseases

Figure 9. Select preparation form and administration route

•

•

the therapeutic principle manually as in Figure 7. Step 5. Input the standard or nonstandard WM diseases. ◦âŠ¦ Step 5.1. Standard WM diseases Check the option Use standard WM disease. Select a record from the WM diseases database as the default value. If there is no corresponding record for certain standard disease in the database or the information isn’t correct, you can add or modify it in the WM disease Maintenance module. ◦âŠ¦ Step 5.2. Nonstandard WM diseases Uncheck the option Use standard WM disease. Input the WM diseases manually. As several diseases may have the same syndromes, users can fill in multiple diseases separated by comma as in Figure 8. Step 6. Select the default preparation forms and default route of administration. The preparation forms might be injection, ene-

•

ma, decoction, etc. The administration route could be by intravenous injection, by enema, by mouse, etc as in Figure 9. Step 7. Update the first-visit symptoms and return-visit symptoms. Click the button First visit symptoms or Return visit symptoms to switch the visit mode. Click the button Add to pop up the symptoms selection dialog. If there is no corresponding record in the database or the information isn’t correct, you can add or modify it in the Symptom Maintenance module as in Figure 10. Users can resort the symptom list by dragging and dropping the symptom items in the list as in Figure 11. The Symptom Maintenance module may be modified sometimes. Steps of modification are further illustrated. ◦âŠ¦ Step A. Select the entry Symptom Maintenance in the module tree to go to its main page. Click button Add to pop up the New CMF dialog.

39


Figure 10. Select the symptoms

Figure 11. Resort the personalized symptom list

◦âŠ¦

◦âŠ¦

40

Step B. Select the root symptom. If you are adding/editing a nonstandard symptom, select the corresponding specialist under other symptoms (e.g. Others → Xiaosun Cai) as in Figure 12. Step C. To describe how serious this symptom is, you can set a question for users to select. The question could be a checkbox, single choice question, multiple choices question, or freeform text question.

•

•

Step 8. Input the examination items. Click the button Examination items and click Add to pop up the examination item selection dialog. If there is no corresponding record in the database or the information isn’t correct, you can add or modify it in the Examination Item Maintenance module. Step 9. Contact the administrator after adding a new template. The administrator will update the backend system. Besides symptoms, syndromes and diseases, CMF is


Figure 12. Add the personalized symptom

Figure 13. Select or edit the herbs

also designed in a personalized way. There are three steps for CMF maintenance. ◦âŠ¦ Step A. Select the entry CMF Maintenance in the module tree to go to its main page. Click button Add to pop up the New CMF dialog.

◦âŠ¦ ◦âŠ¦

Step B. Input the CMF number and name. Step C. Edit drugs. Click the button Add drug. The drug selection dialog will pop up. Check/uncheck the option Standard drug database to

41


Figure 14. Input the basic information of a patient

switch between standard and nonstandard drug database. Users can search the drug by name, ID or name abbreviation. Double click certain drug or click the button Add to append it into the drug list. If certain drug isn’t in this database, you can add it in the Drug Maintenance module as in Figure 13. Users can modify an existing CMF. Resorting the drug list is quite easy by dragging and dropping the drug item in the list.

Demonstration of the Personalized System As the size of the computer screen is limited, the interface of the personalized system for Cai’s Gynecology is shown from Figure 14, figure 15, figure 16, and Figure 17. Main elements in the interface are marked in the four figures. The setup platform for TCM can be visited by the homepage 1, with login name ys and password ys123. The interface is demonstrated in Figure 14. The model tree in the left of the figure lists

42

three main functions of the platform, clinical records input, data management and basic data maintenance. When entering a clinical record, basic information of the patient, cardinal symptoms, accompanied symptoms, results of needed examinations, syndrome differentiator, therapeutic principles, TCM and WM diagnosis, CMF and some attachments should be all included. Nine kinds of basic information maintenance, such as medicine, CMF, syndrome, and so on, are supported in the platform.

A Case: Womb Tumor Tumor is a great challenge to the current medical world. Womb tumor is usually benign and located in the womb, which -–frequently appears in female reproduction system. Womb tumor is classified, according to its location, as intramurals tumor, subserosals tumor, submucosals tumor and cervices tumor caused by the cell growth of smooth muscle in uterus. If there is more than one location, it is multiple-womb tumor. Womb tumor is stony uterine mass in TCM, which takes place in


Figure 15. Select the cardinal symptoms

Figure 16. Select the accompanied symptoms

women whose ages are from 30 to 50, and it will wither after menopause. The symptoms are different depending on the location, size, number and growth speed. Some women feel well, but find it when they conduct physical examination. Many people are with symptoms, e.g. large quantity of menstrual flow and even incessant extramenstral vaginal bleed-

ing. It may reduce or expand the menstrual period leading to anemia. When the diameter of womb is over 10cm, it will constrict other organs and produce other unexpected symptoms, e.g. frequent maturation, urgent urination, even hard urination and defecation. Most women with womb tumor may be infertility, which may be caused by constricting on oviduct of larger tumor.

43


Figure 17. Select syndrome differentiator and therapeutic principle as well as WM diagnosis

Womb tumor may be related to or caused by estrogen. In TCM, it is caused by qi stagnation and/or blood stasis. In the following conditions, womb tumor may take place: 1) cold assailing the womb after production or menstruation; 2) anger hurting the liver system during menstruation leading to qi stagnation; 3) thinking hurting the spleen leading to blood stasis. The therapeutic principle of womb tumor is HuoXueHuaYu and XiaoJianSanJie by Xiaosun (Cai, Huang and Mo, 2000). In the medical records, we also mine that Xiaosun has a so called period therapy technique, i.e. separating one month to be during menstruation and between menstruations. In the period of between menstruations, Xiaosun often uses the therapeutic principle of HuoXueHuaYuXiaoJian. The fundamental prescription is YunFuLing 12g, GuiZhi 3g, ChiShao 10g, DanPi 10g, Taoren 10g, ZaoJiaoCi 30g, JiuJiaPian 9g, ShiJianChuan 10g, GuiJianYu 20g, HaiZao 12g, EShu 10g. There are 14 -21 days in the between menstruation, one day one dose. If the patient is strong, DaHuang, MangXiao may be added to LiangXueHuaYu and RuanJianSanJie, BaiShu is

44

also added to restrict the violent of the above two herbs. If the patient is weak, DangShen is added to make her strong. HuangYaoZhi, YanDanZi, ShuiZhi, DiBieZhong may be added to strengthen the ability of reducing tumor. During the period of menstruation, the therapeutic principle is HuaYuTiaoJing, where the formula of SiWuTaiChong is employed. It composes of ChaoDangGui 10g, DaShengDi 10g, ChuanQiong 5g, BaiShao 10g, ChaiHu 5g, ZhiXiangFu 10g, and HuaiNiuXi 10g. If there is profuse menstruation, Xiaosun insists on HuaYu, not GuSeZhiXue according to the therapeutic principle of treating the unstopped by unstopping. The formula is that ChaoDangGui 10g, DanShen 6g, ChiShao 10g, BaiShao 10g, ShengPuHuang 30g, XueJie 3g, HuaRuiShi 15g, ShuChuanJun 10g, YiMuCao 10g, XianHeCao 20g, ZhenLingDan 12g. If blooding, add SanQi; else if Qi stagnation, add XianFu; else if abdominal pain, add YanHuSuo; else if congealing cold, add AiYe; else if Qi deficiency, add DangShen, ShengHuangQi. Womb tumor is a frequent disease in current clinical gynecology. Basic idea of Xiaosun is add-


ing/deleting the herbs according to the symptoms combining the physique, illness of the patients. If the patient is young and earlier diagnosed, the therapeutic principle is to fight with the disease. While if the patient is older or deficiency of Qi and blood after long term blooding, the therapeutic principle should be FuZhengHuaYu, not fight. If the patient is menopause, then the therapeutic principle is to push amenorrhea and make the tumor shrink itself. If necessary, KuShen, HanShuiShi, XiaKuCao are employed to purge the liver and clear heat, preventing tumor to be malign. A typical medical record is from a woman whose surname is Wang, 35 years old and married. During physical examination, she found the tumor, it is 4.3cm * 7.8cm * 6.4cm. The patient was afraid of operation and looking for TCM therapy. Her menstruation was usually advanced and profuse. The color of menstruation was gray and mixed with blocks. She was distending breast pain before menstruation. Her waist was soreness. Her tongue fur is thin with purple points in margins of tongue. Her pulse is thin and string-like. She often felt tired. Xiaosun judged this patient be SuYuNeiJie in disease pattern, and decided to use the therapeutic principle of HuoXueHuaYu, RuanJianSanJie. The prescription is GuiZhi 3g, ChiShao 9g, DanPi 9g, YunFuLing 12, TaoRenNi 9g, SanLing 9g, EShu 9g, GuiJianYu 20g, ShuiZhi 4.5g, XiaKuCao 12g, HaiZao 9g. 14 doses. The patient did not feel uncomfortable after the therapy. She was in the period of menstruation, quantity is less. She still felt soreness in waist, with thin slimy tongue fur and thin string-like pulse. Therefore, Xiaosun made the prescription according to the previous formula and asked the patient to use after menstruation. After 6 months, the patient performed physical examination by ultrasonic wave and found that the light points were well-distributed, without blocks or dark liquid area. Her menstruation was regular. She was still healthy after one year.

CONCLUSION AND FUTURE WORK This study proposes a novel platform for setting up personalized systems for TCM masters in China. It helps to preserve the specialists’ thoughts and experience in unfiltered ways. The collected materials are invaluable data source for further research. In order to illustrate the idea of the proposed personalized system, clinical records from Mr Xiaosun Cai, a famous national TCM master in gynecologist, are introduced as an example. A complete specialized template for diagnosis is summarized and a personalized system is set up. Data mining methods are also implemented for case analysis. Future work mainly focus on more clinical cases and more types of machine learning methods for data mining. Relationships between symptom and syndrome, symptom and formula will be the focal points.

ACKNOWLEDGMENT Thanks to Xue-Qiang Zeng, Hao-Hua Meng, Lei Ge, Bo Chen and Qing-Ce Zhao for their contributions to the proposed system. This work was supported by the Natural Science Foundation of China under grant no. 60873129, 30901897, the STCSM ”Innovation Action Plan” Project of China under grant no. 07DZ19726, the Shanghai RisingStar Program under grant no. 08QA1403200, the Open Projects Program of National Laboratory of Pattern Recognition.

REFERENCES Cai, X., Huang, S. Y., & Mo, X. H. Y. (2000). Cai Xiaosun’s talk on gynecology. Shanghai, China: Shanghai Education Press of Science and Technology.

45


Chen, H., Wu, Z., Huang, C., & Xu, J. (2003). TCM-Grid: Weaving a medical grid for traditional chinese medicine. In Proceedings of International Conference on Computational Science, (LCNS 2659, pp. 1143–1152). Springer. Honda, K., & Jacobson, J. (2005). Use of complementary and alternative medicine among United States adults: The influences of personality, coping strategies, and social support. Preventive Med, 40(1), 46ł53. Li, G.-Z. (2010). Machine learning for clinical data processing. In Daskalaki, A. (Ed.), Digital forensics for the health sciences: Applications in practice. Hershey, PA: IGI Global. Qin, Z., Mao, Z., & Deng, Z. (2007). The application of rough set in the Chinese medicine rheumatic arthritis diagnosis. Chin J Biomed Eng, 20(4), 357–363. Thomas, K., Nicholl, J., & Coleman, P. (2001). Use and expenditure on complementary medicine in England: A population based survey. Comp Ther Med, 9, 2–11. doi:10.1054/ctim.2000.0407 Wang, X., Qu, H., Liu, P., & Cheng, Y. (2004). A self-learning expert system for diagnosis in Traditional Chinese Medicine. Expert Systems with Applications, 26(4), 557–566. doi:10.1016/j. eswa.2003.10.004 Yamashita, H., Tsukayama, H., & Sugishita, C. (2002). Popularity of complementary and alternative medicine in Japan: A telephone survey. Comp Ther Med, 10, 84–93. doi:10.1054/ctim.2002.0519 Yao, M., Ai, L., Yuan, Y., & Qiao, Y. (2002). Analysis of the association rule in the composition of the TCM formulas for diabetes. Beijing Univ Tradit Chin Med, 25(6), 48–50.

46

You, M., Ge, L., Li, G.-Z., Xu, L., & Huang, S. (2009). A TCM platform for maters’ experience sharing. In Proceedings of the International Joint Conference on Bioinformatics, Systems Biology and Intelligent Computing (IJCBS09) (pp. 388391). You, M., Li, G.-Z., Zeng, X.-Q., Ge, L., Bi, L., & Huang, S. (2008). A personalized Traditional Chinese Medicine system in the case of Cai’s Gynecology. International Journal of Functional Informatics & Personallized Medicine, 1(4), 419–438. doi:10.1504/IJFIPM.2008.022157 Zhang, R., Wang, Y., Zhou, X., Liu, B., Yao, N., & Li, P. (2008). Researching into the experience of TCM veteran practitioners and designing the platform for intelligent data mining. Science and Technique of World: Modernization of Chinese Traditional Medicine, 10(1), 45–53. Zhou, C. (2004). Problems and thinking in the process of experience summarization and carrying forward for TCM veteran practioner, Jiangxi Chinese and Western Medicine Magazine, 25(12). Zhou, X., Wu, Z., Yin, A., Wu, L., Fan, W., & Zhang, R. (2004). Ontology development for unified Traditional Chinese Medical language system. Artificial Intelligence in Medicine, 32(1), 15–27. doi:10.1016/j.artmed.2004.01.014 Zhou, X.-Z., Wu, Z.-H., & Lu, W. (2001). TCMMDB: A distributed multi-database query system and its key technique implementation. In Proceedings of IEEE SMC (pp. 1095–1100). IEEE Computer Society. Zhu, L., Xue, H., Lin, S., Zha, Q., Zhang, Q., & Lv, A. (2007). The association analysis in 445 case of famous practitioner of Chinese medicine about asthma. Jiang Xi University of TCM, 19(5), 83–87.


KEY TERMS AND DEFINITIONS Traditional Chinese Medicine:Traditional medicine in China, which is subject to study human physiology and pathology as well as diagnosis and preventing of disease. It is the accumulation of the theory and experience to fight with disease of the Chinese, whose principle is to view the body in total. A well known idea is to prevent the disease from sprouting by observing their fore gleam. Gynecology: Refers to a subject to study the physiology and pathology of the women reproductive system outside the pregnant period and their diagnosis and therapy. Gynecology usually consists of fundamental knowledge, inflammation, tumor, endocrine, damage, abnormal growth of the reproductive system and other related disease. Cai’s TCM Gynecology: Famous in China, which started 250 years ago in Jiangwan Shanghai. Now the seventh successor is Prof. Xiaosun Cai, who is now 87 years old. Cai’s gynecology takes care with all the symptoms of the patients, employs comprehensive consideration of four diagnostic methods. Their prescription is light and effective. Their successors are asked to be careful and amenable with the patients. Womb Tumor: Takes place frequently in the female reproduction system, which is classified

according to its location as intramurals tumor, subserosals tumor, submucosals tumor and cervices tumor caused by the cell growth of smooth muscle in uterus. If there are more than one location, it is multiple womb tumor. Womb tumor is stony uterine mass in Traditional Chinese Medicine, which takes place in the women whose ages are between 30-50, and withers after menopause. Experience Sharing: A special topic in traditional Chinese Medicine (TCM). Famous TCM practitioners accumulate ample experience integrating the basic theory, the previous experience and their clinical practice. They have the ability to effectively solve doubtful, hard disease like tumor, so they represents the current academic and development of the highest level, which are shown in their medical records and clinical experience. Their experience is hard to be explained in the book, or readers feel hard to understand them even in books. So experience sharing is important in TCM.

ENDNOTE 1

http://levis.tongji.edu.cn/medweb/

47

Section 2

Basic Research:

A Bridge to Modern Medicine

49

Chapter 3

Stem Cell-Based Personalized Medicine: From Disease Modeling to Clinical Applications Alessandro Prigione Max Planck Institute for Molecular Genetics, Germany

ABSTRACT Regenerative medicine is a rapidly evolving research field whose main aims are to provide new therapeutic approaches and to repair or replace injured tissues with functional cells derived from stem cells. In the past few years, research breakthroughs have revolutionized the field by showing that all somatic cells have the potential to re-acquire stem cell-like properties. Thus, it appears possible to generate relevant cell types starting from cells easily obtained from affected individuals. The obtained differentiated cells could eventually serve as in vitro tools for the study of disease-associated mechanisms and for performing customized drug screenings. Moreover, in the context of cellular transplantation, these cells represent the ideal cell source given that they posses the same genetic code and thus will avoid the occurrence of unwanted immune reactions. Overall, this revolutionary technique called cellular reprogramming might provide substantial support for the future development of personalized medicine. In this chapter, I describe the recent advances in the field of stem cell-based regenerative medicine applications. Parkinson’s disease is chosen as a paradigmatic example in which the use of stem cells for study and therapy could have a relevant impact and potentially represent a future cure for this debilitating disorder.

INTRODUCTION Regenerative medicine aims at repairing or replacing lost or injured tissues and organs due diseases, aging, or congenital defects with living and functional cells derived from stem cells. Stem cells are a special cell type that can be found in

every multi-cellular organism. Their unique features include two key properties, i.e. indefinite propagation (self-renew) and the capability to generate a diverse range of specialized cell types through differentiation (potency). Stem cells are classified according to their degree of potency. Embryonic stem cells (ESCs), derived from early-

DOI: 10.4018/978-1-61350-120-7.ch003


Stem Cell-Based Personalized Medicine

stage embryos are pluripotent, i.e. they retain the ability to differentiate into virtually any cell type of the body derived from any of the three germ layers. On the other hand, somatic stem cells, which are found in developed organisms, are usually multipotent, since their differentiation is restricted to one specific lineage. Regenerative-based applications are historically divided into two branches. The first utilizes cellular transplantation of ex vivo-cultured stem cells, while the second seek to activate the endogenous stem cell population. The latter route appear less achievable in the near future, as it requires complex chain of events capable of selectively stimulate the resident stem cell population without inducing unwanted unspecific responses, which might lead to cancerogenic transformations. Thus, most of research studies are currently focused on in vitro stem cell cultures and on the optimization of distinct protocols for the generation of specialized cell types, which could be then used for cellular transplantation. To this end, ESCs represent the most plastic type of stem cells. ESCs were first discovered in mice by Martin Evans in 1981 (Evans & Kaufman, 1981), who has been awarded the Nobel Prize for his landmark discovery in 2007. The derivation of human ESCs was reported by James Thomson in 1998 (Thomson, et al., 1998). Since then, the field has rapidly evolved and considerable advances have been made to define specific differentiation protocols able to efficiently generate functional relevant cell types. These include cardiomyocytes and hepatocytes, which might prove to be useful for drug-screening studies, and dopamine-producing neurons, which could be used in cell-based therapeutic applications aiming at replacing the loss of this specific neuronal population in patients affected by Parkinson’s disease. Despite all these promising results, numerous drawbacks are associated with ESC-mediated regenerative medicine. Indeed, the derivation of human ESCs requires the use of embryos, which can raise ethical concern. Moreover, in a similar

50

fashion to organ transplants, stem cell transplants can give rise to immune rejections, since they are derived from a different individual with a different genetic background (allogenic). Patients should then be subject to immuno-suppressive therapy, which may in turn produce several clinical complications.

Generation of Patient-Derived Pluripotent Stem Cells One way to circumvent the issues related to the use of human ESCs for disease treatment is the generation of genetically equivalent stem cells, directly obtained from the own cells of the patients (isogenic). All the cells of multi-cellular organisms possess the same genetic code but are functionally heterogeneous, due to their distinct epigenetic status. Through the modulating of their epigenetic profile, differentiated somatic cells can be reverted to stem cell-like cells capable of acquiring the two key properties of ESCs, i.e. self-renewal and pluripotency.

Somatic Cell Nuclear Transfer (SCNT) The first demonstration of this reversion was obtained by somatic cell nuclear transfer (SCNT), which removes the nucleus of a somatic cell and transplants it into an enucleated oocyte. In the late 1950s, John Gurdon transferred nuclei from adult frog cells into frog eggs and showed for the first time that the resulting cells took on embryonic characteristics (Gurdon, 1962). This established that, although the body’s cell types retain their genomes as they specialize, it may be possible to re-activate genes that have become functionally inactive during development. The technique was then used by Ian Wilmut in 1997 to generate dolly the sheep. Finally, the derivation of cloned primates has been recently reported. The SCNT approach has been reduced to practice in mouse models to treat genetic immunodeficiency and Parkinson’s disease (Tabar, et


al., 2008). Nonetheless, the derivation of genetically identical “customized” ESC-like cells by SNCT using donor cells directly obtained from patients is rather complicated. Furthermore, it may encounter similar ethical difficulties as human ESC-based medicine, as it also requires the use of human oocytes. Perhaps most importantly, SNCT has yet to be proven successful in the context of human cells.

that iPSCs might be obtained in the future using only a cocktail of different drugs. Overall, cellular reprogramming appears as a procedure that could work for virtually any cell type of the body and does not require the use of human embryos. These features put forward patient-derived iPSCs as the most promising tool for regenerative medicine applications (Kiskinis & Eggan, 2010).

Induced Pluripotent Stem Cells (iPSCs)

iPSCs for Cell Replacement Therapy

Recently, a second route to revert somatic cells to pluripotency has been discovered. This groundbreaking technique, called cellular reprogramming, was first demonstrated by Shinya Yamanka in 2006. Cellular reprogramming consists in introducing ESC-specific transcription factors into somatic cells, which subsequently acquire ESC properties and turn into induced pluripotent stem cells (iPSCs). iPSCs and ESCs share the same cardinal features of self-renewal and pluripotency. iPSCs and ESCs appear almost undistinguishable in terms of gene expression, morphology, cell-cycle structure, epigenetic signature, mitochondrial properties, and, most importantly, developmental potential (Prigione, et al., 2010; Takahashi, et al., 2007; Takahashi & Yamanaka, 2006). The somatic cell source used for cellular reprogramming was first characterized by fibroblasts, but recent studies showed similar results also using different cell types. The procedure was first demonstrated in mouse and later on in humans and other mammals. The methods by witch the factors are introduced into the cells have been also improved. Initially, the technique was dependent on viral transduction, which can result in unwanted genome integration. However, non-integrative methods for iPSC derivation have been recently demonstrated, including the use of plasmids, proteins, RNAs, or microRNAs. Moreover, chemical compounds enhancing the procedure have been identified. Thus, it may be possible to envision

Shortly after the first report describing the methodology of cellular reprogramming, therapeutic potentials of iPSC-based regenerative medicine have been shown by a proof-of-principle study in a humanized mouse model of sickle cell anemia (Hanna, et al., 2007). In the study, fibroblasts derived from an anemic mouse were first reprogrammed to iPSCs and then subjected the cells to gene therapy-based correction of the defective mutant gene. The iPSCs were subsequently differentiated into hematopoietic progenitor cells, which have been transplanted back into the same mouse. As a result of the treatment, pathologic features of the diseases were substantially improved. A similar approach was taken with human fibroblasts obtained from an individual affected by Fanconi Anemia, a blood disorder characterized by genetic instability (Raya, et al., 2009). In this case, the mutant gene was replaced prior to cellular reprogramming. iPSCs could be generated and afterward differentiated into hematopoietic progenitors which maintained a disease-free phenotype. Mouse iPSCs have been also successfully differentiated into functional dopaminergic neurons. These neurons, once transplanted into a rat model of Parkinson’s disease, were able to functionally integrate into the adult brain and in turn lead to a significant amelioration of the disease phenotype (Wernig, et al., 2008).

51


These studies illustrated the potential of iPSC-based regenerative medicine. By inducing somatic cells to re-acquire ESC-like features, cellular reprogramming appears as the holy grail of cellular replacement therapy. It does not involve any ethical problems and, since the transplanted cells can be obtained from the same individual, it avoids the occurrence of unwanted immunological reactions. Thus, iPSC technology offers the unique opportunity to repair damaged cells with genetically equivalent cells. Nevertheless, substantial challenges still remain to be overcome before translating the technique to clinical practice. Factors used for cellular reprogramming are known oncogenes, the non-integrative delivery methods display very low efficiency, and iPSC-derived cells may differentiate less efficiently to relevant cell types when compared to original ESCs. For safe iPSC applications, it will be crucial avoid the presence of undifferentiated cells within the cells prepared for transplantation, as they may give rise to tumors. Finally, significant improvement has to be achieved for the correct delivery of cells into affected individuals and selective functional engraftment of these cells into the corresponding tissue. Overall, further work is warranted toward generating and fully characterizing “clinical grade” iPSCs before attempting their use in human cell replacement therapy. Finally, iPSCs have been recently found to harbor nuclear and mitochondrial genomic alterations, which suggests that reprogramming may be associated with genomic instability (Mayshar, et al., 2010; Gore, et al., 2011; Prigione, et al., 2011). Hence, it will be essential to develop reprogramming protocols able to safeguard the genome integrity of somatic cells, before employing iPSC-based cellular therapy.

52

iPSCs for Disease Modelling Although safe and efficient iPSC-based cell replacement therapy may not be immediately achievable, the use of reprogrammed cells for in vitro modeling of complex disorders appears as a highly useful short-term application of cellular reprogramming. The derivation of specific cell types from iPSCs obtained from patients harbouring genetic mutations or with idiopathic disorders represents indeed an invaluable tool for the study of specific disease-related mechanisms. The concept at the base of utilizing iPSCs to model human disease in a dish is that these cells exhibit the unique capacity to continuously replicate without senescence. Thus, they could represent a limitless reservoir of cells that can be differentiated into virtually any given cell type. This is particularly relevant in the context of human disorders in which the affected cell type can not be obtained and used for biological studies. In neurodegenerative diseases, for example, selective neuronal populations are targeted and lost during time. However, it is technically challenging and ethically impossible to take these neurons from the patients and bring them to a culture dish in order to analyze the mechanistic reasons underlying their selective cell death. While animal models have been crucial for our understanding of disease-related mechanisms, fundamental biological differences exist between mice and humans. A large number of failed clinical trials may indeed be due to these specie-specific differences. In order to better understand the disease pathogenesis and consequently discover and test new disease-modifying drugs, in vitro disease models conducted on human cells are thus necessary. At present culture models are limited to tumor cell lines or transformed derivatives of native tissues, both differing from a “normal” cell due


to immortalization. On the other hand, the novel reprogramming strategy holds the potentiality to generate patient or disease-specific pluripotent cells without the need for human ESCs. Generation of iPS cells from patients with a variety of diseases has been already reported (Park, et al., 2008). These include neurodegenerative disorders, such as Parkinson’s disease (PD), amyotrophic lateral sclerosis (ALS), spinal muscular atrophy (SMA), and Huntington’s disease (HD), and inherited genetic diseases such as adenosine deaminase deficiency-related severe combined immune deficiency (ADA-SCID), ShwachmanBodian-Diamond syndrome (SBDS), Gaucher disease (GD) type III, Duchenne (DMD), Becker muscular dystrophy (BMD), juvenile-onset, type 1 diabetes mellitus (JDM), Down syndrome (DS)/ trisomy 21, the carrier state of Lesch-Nyhan syndrome, and Fanconi anaemia. Effective disease modeling has been successfully demonstrated in two cases. In the first study, iPSCs were generated from a genetic form of spinal muscular atrophy (SMA) and thereafter differentiated into motor neurons, which are the cell type affected by this disorder (Ebert, et al., 2009). The derived neurons were initially similar to those obtained from wild-type iPSCs. However, in depth analysis revealed that their number and size declined with time at faster pace in comparison to the control cells. Most importantly, the cells exhibited protein aggregates which are a characteristic phenotype of SMA. In the second study, iPSCs were derived from patients affected by Familial Dysautonomia (FD), a neurological disease caused by a genetic mutation. Neurons differentiated from the disease-derived iPSCs recapitulate key FD features, which were meliorated by using a specific drug treatment (Lee, et al., 2009). Overall, these examples show the potential of iPSC-based in vitro disease modeling. However, it may be more challenging to in vitro recapitulate diseases with a long latency, such as Alzheimer’s and Parkinson’s disease. In these cases, the dynamics of disease progression in the patient may

be very different form phenotypes developing in vitro in cells differentiated from patient-specific iPSCs. It will thus be necessary to attempt to accelerate the appearance of pathological phenotypes by exposing the cells to certain environmental stressors known to play a role in the development of the disease.

iPSCs for Drug Discovery The revolutionary iPSC technology could immediately be of tremendous help not only for disease modeling but also for another clinically relevant application, i.e. drug discovery. Indeed, by generating hepatocytes and cardiomyocytes from patient-derived iPSCs, it could be possible to perform personalized drug screenings and develop patient-specific drugs. Human iPSCs could then be used to test mechanisms of action and toxicity of potential new drugs as well as drug metabolism and clearance. At present, cardio-toxicity is one of the major forms of drug toxicity and accounts for most recalls and delays experienced in drug development and approvals. A big challenge is how to predict cardiac risks early and accurately in the drug development process. Besides cardio-toxicity, hepatic toxicity is another important complication: distinct patients have different levels and tolerance to the clearance of drugs by their liver, which results in distinct effects of a specified drug on these individuals. Overall, hepato-toxicity and cardio-toxicity are two principle causes of drug failure during pre-clinical testing. Moreover, an additional major problem is represented by the occurring of high variability among individual responses to potential therapeutic agents. Functional cardiomyocytes and hepatocytes have been already efficiently differentiated from mouse and human iPSCs. These cells will soon provide an important in vitro model for screening toxicity of drugs and the development of pharmaceuticals. Currently, cell lines and animal models are utilized for toxicity screening and determining

53


cardiac safety. By using iPSC-derived cells, it will be possible to generate a library of cell lines that may be used to represent a broad spectrum of the population harbouring distinct genetic and epigenetic backgrounds. High-throughput screening might then be used to allow a better prediction of the possible toxicological responses to newly developed drugs. Furthermore, iPSC-derived cells may provide insight into likely mechanisms at the base of the identified toxicity. Overall, this approach might eventually reduce the risks and costs commonly associated with early-stage clinical trials and lead toward a more personalized approach to drug administration.

The Example of Parkinson’s Disease Parkinson’s disease (PD) is the most common movement disorders and second most common neurodegenerative disorder after Alzheimer’s disease, affecting at least 1% of people by age 70 in the western world. The core clinical features, described for the first time by James Parkinson in his classic 1817 monograph and subsequently refined by Jean-Martin Carchot in the last part of the nineteenth century, include tremor at rest, slowness of movement and muscular rigidity (Savitt, Dawson, & Dawson, 2006). Pathological studies determined that the central defect underlying the triad of motor systems resides in the degeneration of dopamine (DA) producing neurons in the substantia nigra pars compacta (SNpc). Affected SNpc neurons typically develop intracytoplasmic inclusions, known as Lewy bodies, which are mainly composed by alpha-synuclein (aSN). Subsequent reduction in the nigrostriatal DA innervation to the striatum deregulates the striatal neuron activity leading to locomotor dysfunctions. Nigral neurons physiologically stimulate the direct pathway and inhibit the indirect pathway of the basal ganglia circuitry,

54

overall facilitating motor movements through the stimulation of the motor cortex. Thus, the loss of SNpc cells results in a hypokinetic movement disorder.

Current Therapeutic Strategies of PD Parkinson’s disease is a chronic disorder that requires broad-based management including patient and family education, general wellness maintenance and nutrition. Current medical and surgical therapies for PD are only symptomatic and lack significant disease-modifying effects. Indeed, after more than thirty years from its introduction, the most effective medical treatment is still based on the DA precursor levodopa, which is mixed with a peripheral decarboxylase inhibitor in order to increase the amount of the drug able to reach the central nervous system (CNS). Additional therapeutic strategies aimed at limiting the breakdown of endogenous DA, through the inhibition of monoamine oxydase type B (MAOB), and at stimulating directly postsynaptic DA receptors, bypassing the need for viable dopaminergic cells to synthesize DA from its precursor. Surgical Deep Brain Stimulation (DBS) appeared as an influential development in symptomatic PD therapy. The results can in fact be quite marked, but the procedure is currently reserved for advanced cases of PD in which motor complications or medication intolerance have led to an unacceptable decline in quality of life. The search for disease-modifying compounds is also an active area of clinical research. Possible neuroprotective strategies include MAOB inhibitors, the antioxidant and mitochondrial component coenzyme CoQ10, and glial-derived neurotrophic factor (GDNF). Finally, one step beyond neuroprotection is cell replacement therapy, which aims to selectively replace the cells lost upon the development of PD.


Cellular Replacement Therapy for PD Although medical or surgical treatments can provide long-term symptomatic relief, the procedures do not slow down disease progression and are haunted by the rise of increasing side effects over time. Hence, there is urgent need for therapeutic strategies that could be reparative in nature and be able to finally provide a cure for this debilitating disorder. Among the most promising is cell transplantation, which holds the hope of repairing the actual neuronal damage in PD rather then merely alleviating the symptoms. The first clinical experiences with cellular therapeutics for PD have been based on fetusderived cells obtained from post-mortem human embryos. The trials were carried out in Sweden in the early nineties by transplanting aborted fetal brain tissues into the brains of PD patients (Lindvall, et al., 1990). The transferred tissues harbored immature DA neurons that were able to successfully integrate into the striatum and produce DA. However, the initial optimism faded when subsequent clinical trials demonstrated decreased efficiency and the occurrence of dyskinesias, or abnormal involuntary movements. Furthermore, recent reports of long-term trials of fetal dopamine grafts revealed the presence of Lewy bodies and phosphorylated aSN within grafted neurons (Li, et al., 2008). Thus, the surviving DA neurons can develop the same pathological features associated with PD, suggesting the spreading of PD to some of the transplanted cells in a similar fashion to host DA neurons within SNpc. The mechanism underlying this spreading remains obscure. However, it appears likely that the unfavorable environment and the immune rejection, commonly associated with allogenic transplantation, may play a critical role. Overall, the use of fetal tissue as a cell source for clinical transplantation appears not very practical. From the start, abortion foes objected the use of fetal-derived tissues. In addition, the number of functional DA neurons obtained from these tis-

sues was very low. The need to generate neuronal cells to be transplanted in large quantities and in a reproducible, steady and safe manner brought the field to consider whether stem cells could offer a feasible alternative. Unlike their fetal counterpart, ESCs are able to produce a large number of functionally active DA neurons in a more reliable and reproducible way. In addition somatic neural stem cells (NSCs), which are multipotent and can give rise to both glial and neuronal cells, could also be used. Both stem cell types, with their capacity for long-term expansion in vitro and their extensive functional stability and plasticity, allowed the establishment of cultures of mature neuronal cells and emerged as amenable cell sources for neural transplantation. Accordingly, transplanted DA neurons derived from murine ESCs were able to restore function in rodent models of PD (Kim, et al., 2002). More recently, the SCNT technique has been used to generate DA neurons using murine fibroblasts as donor cells (Tabar, et al., 2008). The derived DA neurons were transplanted back in the original donor mice which showed significant functional recovery. These encouraging results imply that autologous transplants composed by genetically matched cells could bring fundamental improvement and reduce the problems related to immune rejection. However, as SCNT has not proven successfully for humans, alternative methods able to generate isogenic cell source for transplantation have to be used. To this aim, cellular reprogramming and iPSCs appear as the most promising source for cellular therapy of PD.

iPSCs for In Vitro Modeling and Cellular Therapy of PD The generation of iPSCs cells directly from patient fibroblasts and the consequent derivation of DA neurons might represent a breakthrough in PD research. In the context of in vitro disease modeling, being the closest mimic of in vivo conditions available today, iPSC technology would allow

55


for a more reliable definition of PD neuronal cell biology and cell reactivity. Moreover, it would open the field of regenerative medicine to new exciting venues as the use of DA neurons derived directly from patients may provide a solution to the problem of immune rejection taunting current cell-replacement therapies. Nevertheless, many road-blocks still lay ahead for both iPSC-based applications in PD. First, efficient differentiation protocols generating homogeneous DA neurons have to be established. At present, although several protocols exist, the final population appears composed of mixed cell types. Cellular-sorting technique may provide to be useful for improving the selection of DA neurons. This is important for the study of pathophysiological mechanisms of the disease, which occurs specifically in DA neurons within SNpc. In addition, efficient differentiation and elimination of unwanted undifferentiated cells is highly relevant also for the application of iPSC-mediated cellular therapy. Indeed, the formation of tumors represents as a major problem of cell replacement studies. The cause may be due to the persistence of undifferentiated, proliferating cells among the cells used for the transplantation. Thus, reliable methods for obtaining a pure population of DA neurons have to be applied. The second obstacle lays in the occurrence or lack of occurrence of disease-associated phenotype within iPSC-derived DA neurons. In order to apply iPSC technology to in vitro modeling, DA neurons will have to display typical PD phenotypes in the culture dish. iPSCs have been already generated from PD patients fibroblasts and consequently efficiently differentiated into DA neurons (Park, et al., 2008; Soldner, et al., 2009). However, the differentiated cells did not show any difference when compared to DA neurons derived from healthy control iPSCs. Thus, in order to study PD–related mechanisms, certain cell-stressors might have to be used to facilitate disease progression in vitro. For example, cultured cells may be exposed to environmental effects known to play a

56

role in PD, such as oxidative stress. Indeed, recent studies showed that iPSC-derived neurons from PD patients exhibited higher sensitivity to cellular stressors (Nguyen, et al., 2011). On the other hand, the opposite goal is wanted for the purpose of utilizing differentiated DA neurons for cell transplantation, as these cells have to be healthy in order to contribute to clinical improvement of PD. Thus, if iPSC-derived DA neurons would immediately display clear PD phenotypes, it could represent a complication for future cellular therapies. Indeed, as fetal grafts showed disease features ten years after transplantation, it may be possible to propagate the disease from the host environment into the transplanted cells. Thus, it would be auspicial and necessary to use as a cell source for transplants cells that are as healthy and not disease-compromised as possible. Promising results showed that the course of PD might be indeed modified by iPSCmediated transplants in rodent models of PD. In this study, mouse iPSCs were generated from murine fibroblasts, differentiated into DA neurons and transplanted into Parkinsonian rats, which showed restoration of DA function and disease improvement (Wernig, et al., 2008). The final hurdle facing iPSC-based strategies is that PD pathology is not restricted to the loss of DA neurons. Indeed, over the years, it has become evident that the disease also implicates several non-motor symptoms, including neuropsychiatric, autonomic and sensory complications, thus involving multiple parts of central and peripheral nervous systems beside the extra-pyramidal motor system. Thus, different differentiated cells will have to be analyzed. Moreover, it might be warranted to study co-cultures composed by DA neurons and glial cells in order to better dissect the influence of glial modulation and environmental milieu on disease progression. In the context of cell therapy, it could be possible to experiment different strategies. For example, astrocytes may be used for transplants and the effects on synaptic transmission and glutamate modulation may be


monitored. Alternatively, other cell types might be used, in order to eventually restore the original physiologic circuit not only of DA cells but also of other neuronal populations that may be affected by the disorder. Overall, despite all possible pitfalls, the iPSC technology is expected to open new important avenues in PD research both in the context of in vitro disease modeling and cellular therapeutics.

CONCLUSION The present time is exciting for regenerative medicine research. In particular, high hopes reside in the soon-to-be possible personalized medical applications. A novel strategy, called direct reprogramming, recently demonstrated the possibility to generate

pluripotent cells, which can in turn be differentiated into almost any cell type, starting from a simple skin biopsy. Thus, iPSCs generated from fibroblasts from patients with complex diseases can be differentiated into relevant cell types. This will allow the production of large numbers of differentiated cells with the exact genotype of the diseased ones, which can be used for disease modelling and personalized drug discoveries. Finally, iPSCs may represent a remarkable step forward toward for stem cell therapeutics aimed at curing disorders for which no current therapies significantly modify the disease progression. A paradigmatic example is Parkinson’s disease (Figure 1), a chronic progressive neurodegenerative disorder for which no cure still exists, as current medications or surgical procedures can only provide symptomatic relief. Despite all hurdles that need to be overcome, given the scientific efforts

Figure 1. Schematic description of cellular reprogramming-based regenerative medicine applications in the context of Parkinson’s disease

57


and the significant achievements of the past few year, iPSCs technology have the potential to offer successful and feasible approaches for the study of PD, for its drug intervention, and for cellular treatment strategies that could ultimately provide unprecedented relief to the affected individuals. Parkinson’s disease (PD) is a chronic progressive neurodegenerative disorder characterized by the selective loss of nigrostriatal dopaminergic (DA) neurons. Induced pluripotent stem (iPS) cells can be efficiently derived from fibroblast samples obtained through skin biopsy of affected PD individuals. iPSCs can then differentiate into DA neurons, which can in turn be used for several distinct purposes. In vitro applications include modeling of disease-associated mechanisms and personalized toxicology drug screenings. Finally, iPSC-derived DA neurons could be applied to cellular therapeutic strategy as an autologous source for cell transplantation.

REFERENCES Ebert, A. D., Yu, J., Rose, F. F. Jr, Mattis, V. B., Lorson, C. L., & Thomson, J. A. (2009). Induced pluripotent stem cells from a spinal muscular atrophy patient. Nature, 457(7227), 277–280. doi:10.1038/nature07677 Evans, M. J., & Kaufman, M. H. (1981). Establishment in culture of pluripotential cells from mouse embryos. Nature, 292(5819), 154–156. doi:10.1038/292154a0 Gore, A., Li, Z., Fung, H. L., Young, J. E., & Agarwal, S. (2011). Somatic coding mutations in human induced pluripotent stem cells. Nature, 471, 63–67. doi:10.1038/nature09805 Gurdon, J. B. (1962). Adult frogs derived from the nuclei of single somatic cells. Developmental Biology, 4, 256–273. doi:10.1016/00121606(62)90043-X

58

Hanna, J., Wernig, M., Markoulaki, S., Sun, C. W., Meissner, A., & Cassady, J. P. (2007). Treatment of sickle cell anemia mouse model with iPS cells generated from autologous skin. Science, 318(5858), 1920–1923. doi:10.1126/ science.1152092 Kim, J. H., Auerbach, J. M., Rodriguez-Gomez, J. A., Velasco, I., Gavin, D., & Lumelsky, N. (2002). Dopamine neurons derived from embryonic stem cells function in an animal model of Parkinson’s disease. Nature, 418(6893), 50–56. doi:10.1038/ nature00900 Kiskinis, E., & Eggan, K. (2010). Progress toward the clinical application of patient-specific pluripotent stem cells. The Journal of Clinical Investigation, 120(1), 51–59. doi:10.1172/JCI40553 Lee, G., Papapetrou, E. P., Kim, H., Chambers, S. M., Tomishima, M. J., & Fasano, C. A. (2009). Modelling pathogenesis and treatment of familial dysautonomia using patient-specific iPSCs. Nature, 461(7262), 402–406. doi:10.1038/nature08320 Li, J. Y., Englund, E., Holton, J. L., Soulet, D., Hagell, P., & Lees, A. J. (2008). Lewy bodies in grafted neurons in subjects with Parkinson’s disease suggest host-to-graft disease propagation. Nature Medicine, 14(5), 501–503. doi:10.1038/ nm1746 Lindvall, O., Brundin, P., Widner, H., Rehncrona, S., Gustavii, B., & Frackowiak, R. (1990). Grafts of fetal dopamine neurons survive and improve motor function in Parkinson’s disease. Science, 247(4942), 574–577. doi:10.1126/science.2105529 Mayshar, Y., Ben-David, U., Lavon, N., Biancotti, J. C., & Yakir, B. (2010). Identification and classification of chromosomal aberrations in human induced pluripotent stem cells. Cell Stem Cell, 7, 521–531. doi:10.1016/j.stem.2010.07.017


Nguyen, H. N., Byers, B., Cord, B., Shcheglovitov, A., & Byrne, J. (2011). LRRK2 mutant iPSC-derived DA neurons demonstrate increased susceptibility to oxidative stress. Cell Stem Cell, 8, 267–280. doi:10.1016/j.stem.2011.01.013

Tabar, V., Tomishima, M., Panagiotakos, G., Wakayama, S., Menon, J., & Chan, B. (2008). Therapeutic cloning in individual parkinsonian mice. Nature Medicine, 14(4), 379–381. doi:10.1038/nm1732

Park, I. H., Arora, N., Huo, H., Maherali, N., Ahfeldt, T., & Shimamura, A. (2008). Diseasespecific induced pluripotent stem cells. Cell, 134(5), 877–886. doi:10.1016/j.cell.2008.07.041

Takahashi, K., Tanabe, K., Ohnuki, M., Narita, M., Ichisaka, T., & Tomoda, K. (2007). Induction of pluripotent stem cells from adult human fibroblasts by defined factors. Cell, 131(5), 861–872. doi:10.1016/j.cell.2007.11.019

Prigione, A., Fauler, B., Lurz, R., Lehrach, H., & Adjaye, J. (2010). The senescence-related mitochondrial/oxidative stress pathway is repressed in human induced pluripotent stem cells. Stem Cells (Dayton, Ohio), 28(4), 721–733. doi:10.1002/ stem.404 Prigione, A., Lichtner, B., Kuhl, H., Struys, E. A., & Wamelink, M. (2011). Human iPSCs Harbor Homoplasmic and Heteroplasmic Mitochondrial DNA Mutations While Maintaining hESC-Like Metabolic Reprogramming. Stem Cells (Dayton, Ohio), (Jul): 5. doi:.doi:10.1002/stem.683 Raya, A., Rodriguez-Piza, I., Guenechea, G., Vassena, R., Navarro, S., & Barrero, M. J. (2009). Disease-corrected haematopoietic progenitors from Fanconi anaemia induced pluripotent stem cells. Nature, 460(7251), 53–59. doi:10.1038/ nature08129 Savitt, J. M., Dawson, V. L., & Dawson, T. M. (2006). Diagnosis and treatment of Parkinson disease: Molecules to medicine. The Journal of Clinical Investigation, 116(7), 1744–1754. doi:10.1172/JCI29178 Soldner, F., Hockemeyer, D., Beard, C., Gao, Q., Bell, G. W., & Cook, E. G. (2009). Parkinson’s disease patient-derived induced pluripotent stem cells free of viral reprogramming factors. Cell, 136(5), 964–977. doi:10.1016/j.cell.2009.02.013

Takahashi, K., & Yamanaka, S. (2006). Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell, 126(4), 663–676. doi:10.1016/j. cell.2006.07.024 Thomson, J. A., Itskovitz-Eldor, J., Shapiro, S. S., Waknitz, M. A., Swiergiel, J. J., & Marshall, V. S. (1998). Embryonic stem cell lines derived from human blastocysts. Science, 282(5391), 1145–1147. doi:10.1126/science.282.5391.1145 Wernig, M., Zhao, J. P., Pruszak, J., Hedlund, E., Fu, D., & Soldner, F. (2008). Neurons derived from reprogrammed fibroblasts functionally integrate into the fetal brain and improve symptoms of rats with Parkinson’s disease. Proceedings of the National Academy of Sciences of the United States of America, 105(15), 5856–5861. doi:10.1073/ pnas.0801677105

KEY TERMS AND DEFINITIONS Stem Cells: Specialized cell types able to renew themselves through mitotic cell division and differentiate into a diverse range of specialized cell types. Embryonic Stem Cells (ESCs): Stem cells that are isolated from the inner cell mass of blastocysts and retain the ability to differentiate into cells of all the three germ layers (ectoderm, mesoderm, and endoderm).

59


Induced Pluripotent Stem Cells (iPSCs): A novel population obtained from somatic cells through cellular reprogramming and are believed to display features similar to ESCs. Adult Stem Cells: Found in adult tissues and have limited differentiation abilities; Neural Stem Cells (NSCs) for example can differentiate only into neuronal cells and glial cells. Self-Renewal: Ability to multiply indefinitely and undergo numerous cycles of cell division while maintaining an undifferentiated state with unchanged differentiation potential. Potency: Capacity to differentiate into specialized cell types. Different degrees of potency exist. Totipotency: Ability to give rise to embryonic and extra-embryonic cell types (retained only by spores and zygotes). Pluripotency: Ability to differentiate into virtually any cells derived from all the three germ layers (ESCs, iPSCs). Multipotency: Ability to differentiate into a related family of cells (adult stem cells). Unipotency: Ability to produce only one cell type (germline stem cells can only give rise to either sperm or oocytes). Differentiation: Process by which less specialized (undifferentiated) cells become more specialized cell types. Differentiation normally occurs during the development of a multi-cellular organism. Through differentiation, stem cells are able to generate distinct cell types in vitro and in vivo. Cellular Reprogramming: Procedure of generating pluripotent stem cell artificially derived

60

from non-pluripotent somatic cells. It is obtained by inducing the forced expression of certain key genes, associated with self-renewal features and commonly expressed in undifferentiated ESCs. It may allow researchers to obtain pluripotent stem cells, which are important in research and potentially have therapeutic uses, without the controversial use of embryos. In Vitro Disease Modeling: Cellular platform to dissect the biochemical mechanisms and molecular processes which are altered upon disease development. iPSCs may represent the ideal candidates since they can be directly derived from affected individuals and retain the ability to differentiate into virtually any cell type of the body. Cellular Replacement Therapy: Transplantation of stem cell-derived cells into damaged tissue in order to replace or repair diseased or injured host cells. The type of cell source used for transplantation can have an impact on the final outcome. Allogenic Transplantation: Use of a cell source harboring a different genetic background. The major complication of this procedure is the occurrence of immune rejection due to the body response to cells that are not recognized as self. Isogenic Transplantation (Autologous): Use of a cell source harboring the same genetic background as the target individual in which the cells will be transplanted. This procedure is believed to overcome the problems related to immune rejection. iPSC technology may provide to be useful since it can generate patient-specific genetically identical cells.

61

Chapter 4

Bifurcation Analysis of a Model Accounting for the 14-3-3σ Signalling Compartmentalisation Svetoslav Nikolov Institute of Mechanics and Biomechanics, Bulgaria Julio Vera University of Rostock, Germany Olaf Wolkenhauer University of Rostock, Germany

ABSTRACT Bifurcation theory studies the qualitative changes in the phase portrait when we vary the parameters of the system. In this book chapter we adapt and extend a mathematical model accounting for the subcellular localisation of 14-3-3σ, a protein involved in cell cycle arrest and the regulation of apoptosis. The model is analysed with analytical tools coming from Lyapunov-Andronov theory, and our analytical calculations predict that soft (reversible) loss of stability takes place.

INTRODUCTION 14-3-3 proteins have crucial roles in a variety of cellular responses including signal transduction, cell cycle progression, metabolic regulation and apoptosis (Yang, Wen, Chen, Lozano, & Lee, 2003; Wilker, van Vugt, Artim, et al., 2007). Firstly identified through their high level of expression in the mammalian brain, it is currently well-known that mammals have eight different protein forms

of 14-3-3 that are encoded by seven distinct genes (β, ε, γ, η, τ, ξ, σ). One particular 14-3-3 isoform, σ, is a p53-responsive gene and was characterized as a human mammary epithelium-specific marker (that is the reason why 14-3-3σ is also called stratifin). 14-3-3σ has a variety of cell functions. Regarding cancer and tumour progression, 14-3-3σ plays a role like regulator of the cell cycle, probably via cytosolic sequestration of critical cell cycle proteins like cyclin B1 and

DOI: 10.4018/978-1-61350-120-7.ch004


Bifurcation Analysis of a Model Accounting for the 14-3-3σ Signalling Compartmentalisation

cdc2 in response to DNA damage and other stress signals (Hermeking, 2006). Interesting enough, 14-3-3σ can delay the initiation of apoptosis by sequestering mitochondrial p53 in the cytosol and preventing activation of Bax and other initiators of the apoptotic cascade. This assigns to 14-3-3σ a dual role in the regulation of the cell fate after DNA damage (Samuel, Weber, Rauch, Verdoodt, Eppel, McShea, Hermeking & Funk, 2001). The function of 14-3-3σ is frequently lost in human tumours including melanoma, breast and prostate cancers, which suggests that the protein may act as a tumour suppressor. However, the molecular basis for the tumour suppressor function of 14-33σ remains unknown (Wilker, van Vugt, Artim, et al., 2007). Moreover, 14-3-3σ is silenced in many cancers via gene methylation (Ferguson, Evron, Umbrich, & et al., 2000; Iwada, Yamamoto, Sasaki, & et al., 2000; Lodygin, Diebold, & Hermeking, 2004, Schultz et al. 2009). Remarkably, 14-3-3σ is involved in a positive feedback loop with its own activator p53 via stabilization and inhibition of ubiquitination (Yang, Wen, Chen, Lozano & Lee, 2003). With respect to dynamic systems theory (Glass & Mackey, 1988; Shilnikov, Shilnikov, Turaev, & Chua, 2001; Nikolov, 2004), the first Lyapunov value is one of the basic analytical tools to investigate the transition between different dynamical states in biochemical systems. The (un) stability of these transitions may have important consequences on the dynamics of the system, pointing towards changes where parameters are critical for the emergence of pathological configurations. In addition, qualitative knowledge emerging from this analysis could be used in the development of diagnostic methods and in drug discovery (Nikolov, Vera, Kotev, Wolkenhauer, & Petrov, 2008; Nikolov, Vera, Rath, Kolch, & Wolkenhauer, 2009). The present chapter extends our previous results in the mathematical modelling of 14-3-3σ

62

(Vera, Schultz, Ibrahim, Wolkenhauer, & Kunz, 2009). We here adapt and extend the model to consider the subcellular localisation of the protein (cytosol, nucleus and mitochondria). The model is analysed with analytical tools coming from Lyapunov-Andronov theory to investigate the system and perform analytical (qualitative) predictions, which are complemented by numerical simulation.

MATHEMATICAL MODEL Qualitative Analysis For our investigation, we modified and expanded our previously published model (Vera, Schultz, Ibrahim, Wolkenhauer, & Kunz, 2009). The version of the model discussed here considers the synthesis, dimerisation, degradation and translocation of 14-3-3σ, but also the compartmentalisation of the signaling molecule in the cytosol, nucleus and mitochondria. The model is depicted in Figure 1. As we previously indicated, 14-3-3σ expression is mediated by p53 in response to DNA damage signalling and is able to block the progression of the cell cycle in both G1/S and G2/M transitions, but it also may delay/block the activation of apoptosis (Figure 1, left-hand side). In our model we consider the following critical events in the dynamics of 14-3-3σ: i) p53 mediated synthesis; ii) 14-3-3σ homodimerisation/activation in the cytosol; iii) translocation between cytosol and nucleus; iv) translocation between cytosol and mitochondria; and v) cytosolic degradation of 14-3-3σ (Figure 1, right-hand side). Based on these hypotheses, we derived a model in ordinary differential Equations (ODE), which describes the dynamics of 14-3-3σ. The model considers the following variables accounting for the subcellular distribution of the protein: monomeric cytosolic (σc), dimeric cytosolic (σcD),


Figure 1. Scheme of our mathematical model describing the compartmentalisation of 14-3-3σ signalling. The left-hand side figure described the p53-mediated expression of 14-3-3σ in response to DNA damage and the regulation of cell cycle progression and apoptosis by the protein. The right-hand side is the scheme depicting the processes included in the mathematical model.

dimeric nuclear (σnD) and dimeric mitochondrial (σmD). The ODE model derived has the following mathematical structure: d σc = kS (t ) − 2kD σC2 , dt d σcD = kD σC2 − (kcn + kcm + kdeg ) σcD + knc σnD + kmc σmD , dt d σnD = kcn σcD − knc σnD , dt d σmD = kcm σcD − kmc σmD , dt

(1)

where σc, σcD, σnD and σmD are the already defined time-dependent variables describing the subcellular dynamics of 14-3-3σ and ks, kD, kdeg, kcn, knc, kcm, kmc are the rate constants of the model (1), whose values were either extracted from (Vera,

Schultz, Ibrahim, Wolkenhauer, & Kunz, 2009) or assigned to qualitatively describe the dynamics of the system. For instance, kdeg and kcn were directly extracted from our previous model, while for the parameters knc,kcm, kmc we considered an interval of feasible values around the values described in our previous model to investigate how the dynamics of the system is affected by the modulation of the related processes. Finally, dimerisation of 14-3-3σ is assumed to be a fast, efficient, irreversible process and a high value is assigned to the parameter describing it. The chosen values for the parameters are included in Table 1. For our analysis, let k0 = ks ,

k1 = 2kD ,

k 2 = kD ,

k 3 = kcn + kcm + kdeg ,

k4 = knc ,

k5 = kcn ,

k6 = kcm ,

k7 = kmc ,

y1 = σc ,

y 2 = σcD ,

y 3 = σnD ,

y 4 = σmD .

(2)

63


Table 1. Model parameter values ks = 1,

kD = 2,

kcm = kcn ,

k deg = 0.0069,

kmc = X .kcn ,

kcn = 0.2336,

X ∈ [0.1, 3].

Then substituting (2) into (1), we get dy1 dt dy2 dt dy 3 dt dy 4 dt

Hence, after some transformations the system (3) has the form

= k0 − k1y12 , = −k 3y2 + k 4y 3 + k7y 4 + k2y12 ,

(3)

= k 5 y 2 − k 4y 3 , = k6y2 − k7y 4 .

The steady (fixed point in the phase space) state of the system (1) is found by equating the right-hand sides of (3) to zero. It is easy to see that the steady state is only one with coordinates k y1 = 0 , k1

k _ y 2 = 4 y3 , k5 _

−

kk _ y 4 = 4 6 y 3, k5k7 _

k0k2k5 . y3 = k1k 4 (k 3 − k5 − k6 )

(4)

.

∂ yi D4 = ∑ = −2k1y1 − k 3 − k 4 − k7 < 0, i =1 ∂yi (5)

and the system (3) has an attractor. Generally, in order to determine the character of fixed point (Equation (4)) we linearise (3) near equilibrium solution, i.e. _

yi = x i + y i ,

64

(i = 1, ..., 4)

dx 1 dt dx 2 dt dx 3 dt dx 4 dt

= −c1x 1 − k1x 12 , = c2x 1 − k 3x 2 + k 4x 3 + k5x 4 + k2x 12 ,

= k 5 x 2 − k 4x 3 , = k 6x 2 − k 7 x 4 , (7)

where _

c1 = 2k1 y 1 ,

_

c2 = 2k2 y 1 .

(8)

_

The divergence of the flow (3) is n

knc = X .kcn ,

(6)

According to (Bautin, 1984), the Routh-Hurwitz conditions for stability of (4) can be written in the form: (see Box 1) The notations p, q, r, s and R are taken from (Bautin, 1984). When the conditions (12) or (13) are not valid, the steady state (4) becomes unstable. Here we note that conditions (9)-(11) and (13) are always valid in this case. The characteristic Equation of the system (7) (which is equivalent of system (3)) can be written as χ 4 + p χ 3 + q χ 2 + r χ + s = 0.

(14)

The stability of a steady state (4) depends on the real part of the roots of the characteristic Equation (14). If all roots are negative then the equilibrium state is stable. If at least one root is positive, then the steady state is unstable. Ac-


Box 1. p = c1 + k 3 + k4 + k7 > 0,

(9)

q = c1 (k3 + k4 + k7 ) + k3 (k4 + k7 ) + k4k7 − k5 (k4 + k6 ) > 0,

(10)

r = c1 k 4 (k 3 − k5 ) + k 3k7 − k5k6 + k 4k7  + k 4 (k 3k7 − k5k6 − k5k7 ) > 0,

(11)

s = c1k 4 k 3k7 − k5 (k6 + k7 ) > 0,

(12)

R = pqr − sp 2 − r 2 > 0 .

(13)

cording to (Andronov, Witt, & Chaikin, 1966; Bautin, 1984), the conditions R=0 and s=0 are the boundaries of stability. In the boundary of stability s=0, the characteristic Equation (14) has one root equal to zero, and the type of the other roots is determined by the expression Ω = 27r 2 − 18pqr + 4q 3 + 4p 3r − p 2q 2 .

(15)

Thus, we have two cases: 1. If Ω < 0, p > 0, q > 0, r > 0, R > 0 and s=0, then the Equation (14) has one root equal to zero and three negative real roots; 2. Ω > 0, p > 0, q > 0, r > 0, R > 0 and s=0, then the Equation (14) besides one zero root also has a negative root and two complex conjugate roots with negative real parts. For systems with structurally unstable equilibrium states, the stability theory considers various aspects of stability critical cases. Note that here is included also the bifurcation phenomena. Here, we mention only the two most common and simple cases (Shilnikov, Shilnikov, Turaev, & Chua, 2001; Nikolov, Stoytchev, & Bozhov, 2006), where the characteristic Equation (14) (1) has one zero root

and (2) has a pair of complex-conjugated roots on the imaginary axis. The first case is determined by the condition s = 0,

∆k > 0

(k = 1, 2, 3) ,

(16)

where ∆k is the Routh-Hurwitz determinant. Recall that s=(-1)4detA, where A is the matrix of the linearised system at the equilibrium state. In view of this condition, the equilibrium states associated with the first critical case are also called degenerate. Since the implicit function theorem may not longer be applied here, the persistence of such equilibrium state in a neighbouring system is not necessary guaranteed. Thus, a transition through the stability boundary in the first critical case may result in disappearance of the equilibrium state. In this case the system is structurally unstable (un-robust) and through bifurcation it will lose its stability non-reversely. Generally, the stability of cell signalling pathways could, from biological point of view, be connected to homeostasis, i.e. process of keeping an internal environment stable by making adjustments to changes in the external environment. This is achieved by a system of feedback control loops. In other words, for the stability of cell signalling process it is essential that the cell maintains a stable condition where in

65


fact a constant flux of molecules occurs. However, in this case (p53-sigma) pathway, the homeostasis is disturbed. Studies have shown that such interaction has been observed in cancer disease classifies this type of interaction as disruptive and causing disease. The second critical case corresponds to s > 0,

∆n −1 = 0,

∆k > 0,

(n = 4, k = 1, 2) (17)

In this case on the contrary of the first critical case, the equilibrium state is preserved in all nearby systems and cal loss its stability. From a biological point of view, this means that the homeostasis can be disturbed but after a certain period of time it will restore. If cancer disease appears, it can be healed with medical assistance. Further, we focus our considerations on the problem of the transition over the stability bound1−X kcm . Here we note that ary s=0, i.e. kdeg = X for X∈ (0,1), the fourth Routh-Hurwitz condition for stability (Equation (12)) can be negative. This question has an immediate significance for the subject of nonlinear dynamics. For stationary regimes, the corresponding problem was solved in (Bautin, 1984). There the boundaries of stability are classified as safe or dangerous- safe boundaries (soft loss of stability) are such that crossing them leads to only small quantitative changes of the system’s state; dangerous boundaries (hard loss of stability) are such that arbitrarily small perturbations of the system beyond them cause significant and irreversible changes in the system’s behaviour. Generally in accordance with Lyapunov-Andronov theory, the so-called first Lyapunov value l1(λ0) determines the character (safe or dangerous) of the boundary of stability s=0, when bifurcation parameter λ0 is slowly changed. Thus, in order to define the type of stability loss of steady state (4) it is necessary to calculate l1(λ0) on the boundary of stability s=0. In the case of fourth first order differential Equa-

66

tions, this value can be determined analytically by the formula in (Bautin, 1984):  1 (1) 2  (1) 2 (1) 2 (1) 2 l1 (λ0 ) = α  a11 σ1 + a 22 σ2 + a 33 σ3 + 2 a 44 (1 − ασ1 − βσ2 − γσ3 ) +  δ   2 (1) (1) (1) + a14 σ1 + a24 σ2 + a 34 σ3 (1 − ασ1 − βσ2 − γσ3 ) + δ           (1) (1) (1) σ1σ2 + a13 σ1σ3 + a 23 σ2 σ 3 + β  + 2 a12 ... + γ ... + δ ... .    1      2     3  4

(

)

(

)}

(18)

where λ0 is defined as a value of X or kdeg for which the relation s=0 takes place. Here (1) (2) a11 = −k1 , a11 = k2 and all other aijl (i,j,l=1,2,3,4) are equal to zero for the system (7). The coefficients α, β, γ, δ, σ1 , σ2 and σ3 are defined by corresponding formulas presented in (Bautin, 1984). After accomplishing some transformations and algebraic operations for the first Lyapunov value l1(λ0) (for system (8)) we obtain the main result in this article, i.e. l1(λ0)=0. In other words, the boundary of stability s=0 is safe and soft loss of stability take place.

Numerical Analysis In previous section, we introduced the analytical tools proposed and used them for a qualitative analysis of the system obtaining some predictions about the dynamics of the system. The values chosen for the parameters and used in numerical analysis are those in Table 1. In order to compare the predictions with numerical results, the governing Equations of the model, represented by (1), were solved numerically using MATLAB (Mathwork, 2009). In Figures 2 and 3 we illustrate the dependence of the model behaviour on the value for X –. The parameter X accounts for the relation between the two fluxes of subcellular translocation for a given compartment (nucleus or mitochondria). For instance, a value of X smaller than one accounts for conditions in which the translocation cytosol→nucleus is more efficient than the re-


Figure 2. Stable solutions of system (1), when parameter X is equal to 2.5. In this case all Routh-Hurwitz conditions are positive.

verse process nucleus→cytosol; thus, under this biological setup the protein may accumulate in the nucleus. In opposite case, a value for X bigger than one would indicate that the translocation nucleus→cytosol is more efficient and therefore most of the protein will remain in the cytosol. The same notion applies to the processes between cytosol and mitochondria. Since both processes (translocation in and out the given compartment) may be regulated by different proteins and mechanisms, is it meaningful to consider how changes in their interplay may affect the dynamics of the system. In our numerical simulations it is seen that for X equal to 2.5 the system has stable solutions. Note that in this case all Routh-Hurwitz conditions are valid. On the other hand for X=0.35 (see Figure 3) the system is in unstable zone of parameters space and has unstable solutions. According to analytical results from previous chapter, loss of stability is reversible and after drug therapy the system can be again stable.

CONCLUSION In this book chapter we present a study of the dynamical features of a 4D model describing the

compartmentalisation dynamics of 14-3-3σ using the Lyapunov-Andronov bifurcation theory. During the last decade, the robustness has been considered as a key property of the biological (biochemical) systems (Kitano, 2004; Nikolov, Yankulova, Wolkenhauer, & Petrov, 2007). Under this assumption, small changes in the internal or external conditions of a biochemical system are neutralised by the system which is able to return to the vicinity of the preliminary attractor, while intense changes on the values of parameters characterising the system can provoke the transition to a new state, potentially a different attractor with its own interval of robustness (Kitano, 2004). By knowing the sign of the first Lyapunov value at the bifurcation boundary s=0 as we showed in this work, we are able to define the structural stability (robustness) of the steady state. Here we obtain that first Lyapunov value l1(λ0) for 14-3-3σ system is always equal to zero and soft (reversible) loss of stability take place. From a biological perspective, here we investigate the dynamics of subcellular translocation for 14-3-3σ. This protein is synthesised and degraded in the cytosol and has experimentally verified biological activity in the cytosol itself, the nucleus and the mitochondria. Since the translocation in and out a given compartment (in

67


Figure 3. Unstable solutions of system (1), when parameter X is equal to 0.35. In this case only RouthHurwitz condition s is negative, i.e. s=-0.0114.

our case, the nucleus and mitochondria) may be regulated by different proteins and mechanisms, is it important to consider whether the interplay between both translocation processes may affect the dynamics of the system and induce anomalous accumulation of the protein in any compartment. In the normal physiological configuration of the system, experimental evidences suggest that translocation cytosol→nucleus is less efficient than the reverse process nucleus→cytosol (X bigger than one), most of existing amount of 14-3-3σ remains in the cytosol, where it is degraded, and only a reduced fraction of the protein translocates to the nucleus and mitochondria (Hemert, Niemantsverdriet, Schmidt, Backendorf & Spaink, 2004). Our analysis suggest that under this normal conditions the solutions of the system are always stable, which means that there is a dynamical equilibrium between protein synthesis, degradation and translocation. On the other hand, under pathological conditions in which the translocation cytosol→compartment is more efficient that the reverse one (X smaller than one), our analysis suggest that the systems becomes unstable and big amounts of the protein accumulate on time

68

in both compartments, leading potentially to an abnormal modulation of the cell cycle progression and apoptosis. We hypothesised that this abnormal accumulation of the protein may be related to the fact that degradation of the protein in our models occurs only in the cytosol. In the coming future, we will improve the characterisation of our model using specifically designed experiments and consider the effect of further regulatory structures in the system, like the positive feedback loop p53-MDM2-14-3-3σ.

ACKNOWLEDGMENT Julio Vera is funded by the German Federal Ministry of Education and Research (BMBF) as part of the project CALSYS-FORSYS under contract 0315264 (www.sbi.uni-rostock.de/calsys).


REFERENCES Andronov, A., Witt, A., & Chaikin, S. (1966). Theory of oscillations. Reading, MA: AddisonWesley. Bautin, N. (1984). Behaviour of dynamical systems near the boundary of stability. Moscow, Russia: Nauka. Ferguson, A., Evron, E., Umbricht, C., Pandita, T., Chan, T., & Hermeking, H. (2000). High frequency of hypermethylation at the 14-3-3sigma locus leads to gene silencing in breast cancer. Proceedings of the National Academy of Sciences of the United States of America, 97, 6049–6054. doi:10.1073/pnas.100566997 Hemert, M. J., Niemantsverdriet, M., Schmidt, T., Backendorf, C., & Spaink, H. P. (2004). Isoformspecific differences in rapid nucleocytoplasmic shuttling cause distinct subcellular distributions of 14-3-3 sigma and 14-3-3 zeta. Journal of Cell Science, 117(8), 1411–1420. doi:10.1242/jcs.00990 Hermeking, H. (2006). The c-MYC/p53/14-3-3σ network. Zellbiologie Aktuell, 32(3), 10–12. Iwata, N., Yamamoto, H., Sasaki, S., Itoh, F., Suzuki, H., & Kikuchi, T. (2000). Frequent hypermethylation og CpG islands and loss of expression of the 14-3-3sigma gene in human hepatocellular carcinoma. Oncogene, 19, 5298–5302. doi:10.1038/sj.onc.1203898 Kitano, H. (2004). Biological robustness. Nature, 5, 826–837. Levine, A. (1997). p53, the cellular gatekeeper for growth and division. Cell, 88, 323–331. doi:10.1016/S0092-8674(00)81871-1 Lodygin, D., Diebold, J., & Hermeking, H. (2004). Prostate cancer is characterized by epigenetic silencing of 14-3-3sigma expression. Oncogene, 23, 9034–9041. doi:10.1038/sj.onc.1208004

Matlab. (2009). The MathWorks Inc. Natick, MA, USA. Retrieved from www.mathworks.com Nikolov, S. (2004). First Lyapunov value and bifurcation behaviour of specific class threedimensional systems. International Journal of Bifurcation and Chaos in Applied Sciences and Engineering, 14(8), 2811–2823. doi:10.1142/ S0218127404011004 Nikolov, S., Stoytchev, St., & Bozhov, B. (2006). Mathematical model of blood flow pulsations in the circle of Willis. Comptes rendus de l’Academie bulgare des Sciences, 59(8), 831-840. Nikolov, S., Vera, J., Kotev, V., Wolkenhauer, O., & Petrov, V. (2008). Dynamic properties of a delayed protein cross talk model. Bio Systems, 91, 51–68. doi:10.1016/j.biosystems.2007.07.004 Nikolov, S., Vera, J., Rath, O., Kolch, W., & Wolkenhauer, O. (2009). Role of inhibitory proteins as modulators of oscillations in NFkB signalling. IET Systems Biology, 3(2), 59–76. doi:10.1049/iet-syb.2008.0105 Nikolov, S., Yankulova, E., Wolkenhauer, O., & Petrov, V. (2007). Principal difference between stability and structural stability (robustness) as used in systems biology. Nonlinear Dynamics Psychology and Life Sciences, 11(4), 413–433. Samuel, T., Weber, H. O., Rauch, P., Verdoodt, B., Eppel, J. T., & McShea, A. (2001). The G2/M regulator 14-3-3sigma prevents apoptosis through sequestration of Bax. The Journal of Biological Chemistry, 276(48), 45201–45206. doi:10.1074/ jbc.M106427200 Shilnikov, L., Shilnikov, A., Turaev, D., & Chua, L. (2001). Methods of qualitative theory in nonlinear dynamics, part 2. World Scientific. Vera, J., Schultz, J., Ibrahim, S., Wolkenhauer, O., & Kunz, M. (2009). Dynamical effects of epigenetic silencing of 14-3-3σ expression. Molecular BioSystems, 6, 264–273. doi:10.1039/b907863k

69


Wilker, E., van Vugt, M., Artim, S., Huang, P., & Petersen, C. (2007). 14-3-3σ controls mitotic translation to facilitate cytokinesis. Nature, 446, 329–332. doi:10.1038/nature05584 Yang, H., Wen, Y., Chen, C., Lozano, G., & Lee, M. (2003). 14-3-3σ positively regulates p53 and suppresses tumor growth. Molecular and Cellular Biology, 23(20), 7096–7107. doi:10.1128/ MCB.23.20.7096-7107.2003

70

KEY TERMS AND DEFINITIONS Bifurcation Theory: This theory describes qualitative changes in phase spaces that occur as parameters are varied in the definition of a dynamical system. Safe Boundaries: They are such that crossing over them leads to only small quantitative changes of the system state. Dangerous Boundaries: They are such that arbitrary small perturbations of the system beyond them cause significant and irreversible changes in the systems behaviour.

71

Chapter 5

Engineering Gene Control Circuits with Allosteric Ribozymes in Human Cells as a Medicine of the Future Robert Penchovsky Sofia University “St. Kliment Ohridski”, Bulgaria

ABSTRACT Systems and synthetic biology promise to develop new approaches for analysis and design of complex gene expression regulatory networks in living cells with many practical applications to the pharmaceutical and biotech industries. In this chapter the development of novel universal strategies for exogenous control of gene expression is discussed. They are based on designer allosteric ribozymes that can function in the cell. The synthetic riboswitches are obtained by a patented computational procedure that provides fast and accurate modular designs with various Boolean logic functions. The riboswitches can be designed to sense in the cell either the presence or the absence of disease indicative RNA(s) or small molecules, and to switch on or off the gene expression of any exogenous protein. In addition, the riboswitches can be engineered to induce RNA interference or microRNA pathways that can conditionally down regulate the expression of key proteins in the cell. That can prevent a disease’s development. Therefore, the presented synthetic riboswitches can be used as truly universal cellular biosensors. Nowadays, disease indicative RNA(s) can be precisely identified by employing next-generation sequencing technologies with high accuracy. The methods can be employed not only for exogenous control of gene expression but also for re-programming the cell fate, anticancer, and antiviral gene therapies. Such approaches may be employed as potent molecular medicines of the future. DOI: 10.4018/978-1-61350-120-7.ch005


Engineering Gene Control Circuits with Allosteric Ribozymes in Human Cells as a Medicine of the Future

INTRODUCTION In the last several years sequencing technology has been improving dramatically (Reis-Filho, 2009). Nowadays, 200 million bases can be sequenced per day in the Human Genome Project (HGP) using the latest generation sequencing technology such as the Genome Analyzer IIe from Illumina and the Solid System IV from Applied Biosystems (Oetting, 2010). These machines offer ultra-throughput whole genome sequencing (Bentley et al., 2009), targeted re-sequencing, SNP analysis (Chan, 2009; Ramos et al., 2009), gene expression and small RNA analyses (Jung, Hansen, Makunin, Korbie, & Mattick, 2010), chromatin immune-precipitation and DNA methylation arrays. The next generation sequencing (NGS) technology promises to advance genomic research in the next years by producing whole genome and gene expression data in combination with epigenetic data (Liang et al., 2009). Such complex approach was not available only a few years ago. The huge quantity of experiential data produced by the NGS technology can be fully analyzed by parallel computing (Bateman & Quackenbush, 2009). The combination of NGS technology with high-performance computing (HPC) is set to make a breakthrough in genomics, and many have far-reaching applications in cancer research. This integrated approach promises to bring to light the complexity of factors that make one cell become cancerous (Morrissy et al., 2009; Wyman et al., 2009). Understanding the molecular mechanisms of cancerogenesis is an important step for finding a cure. However, knowing the molecular mechanisms of one disease doesn’t lead immediately to its cure. In this chapter we describe two general approaches that can tackle various disorders, which are associated with expression of disease indicative RNA(s). (Teng & Xiao, 2009; Willenbrock et al., 2009). The methods described in this chapter are based on allosteric hammerhead ribozymes (Porta & Lizardi, 1995), which can work as biosensors in vitro and as synthetic riboswitches and gene regula-

72

tory elements in vivo (R. Breaker & Penchovsky, 2008; Penchovsky & Breaker, 2005). Ribozymes are naturally occurring RNA molecules that possess a catalytic function similar to that of protein enzymes (Muller, 2009). Allosteric enzymes are biopolymers that can change their conformation on binding a specific effector molecule(s) to their allosteric domains (Penchovsky & Breaker, 2005). Effector binding domains are distant from the catalytic center in an allosteric enzyme. Allosteric protein enzymes are discovered to play an important role in biochemical pathways of many organisms. Unlike allosteric protein enzymes all allosteric ribozymes are synthetic molecules. The only naturally occurring ribozyme discovered up to now to sense the presence of an effector molecule, is known as the glmS ribozyme (Blount, Puskarz, Penchovsky, & Breaker, 2006). However, the glmS RNA is not an allosteric ribozyme. Its activator glucosamine-6-phosphate serves as a co-factor in the catalytic center of the ribozyme. The allosteric ribozymes are designed to sense the presence of small molecules or oligonucleotides and to catalyze certain chemical reactions by various engineering methods. Such methods include in vitro selection (Guryev & Cuppen, 2009), rational design (Tang & Breaker, 1997), and computational design procedures (Penchovsky & Breaker, 2005). All allosteric ribozymes described in the current chapter are designed by a published computational procedure (Penchovsky & Breaker, 2005), which is also a subject of a pending patent application (R. Breaker & Penchovsky, 2008). The procedure yields a ribozyme sequence within minutes on an average personal computer with over 90% accuracy. It is advantageous in comparison to in vitro selection and rational design methods in terms of design accuracy and time spent. All presented ribozymes have a modular design that allows the allosteric domains to be easily altered. As a result, the ribozymes can be designed to sense any DNA and/or RNA oligonucleotide from 16 up to 22 nucleotides (nt) long. The ribozymes can be engineered to perform essential Boolean logic


functions such as AND, OR, NOT, and YES (Penchovsky & Breaker, 2005). Moreover, allosteric ribozymes can be engineered to execute integrated logic circuits, which will be published elsewhere. The designer ribozyme can be used as biosensors both in vitro and in vivo. When used as biosensors in vitro (in a solution) all allosteric ribozymes are based on the minimal version of the hammerhead ribozyme. These ribozymes can be immobilized in microfluidic devices to build automated microarrays (Penchovsky, Birch-Hirschfeld, & McCaskill, 2000). The designer ribozymes intended to be used in vivo are based on the wild-type ribozyme from the human parasite Schistosoma mansoni. Wild-type schistosoma hammerhead ribozyme has been used for exogenous control of gene expression both in human cell lines and in mouse (Martick, Horan, Noller, & Scott, 2008; Yen et al., 2004). In contrast to the minimal version, the schistosoma hammerhead ribozyme can function under physiologically relevant conditions due to its specific tertiary structure as described in the next section. Therefore, designer constructs based on the schistosoma hammerhead ribozyme are to be employed in all in vivo applications. The oligonucleotide-sensing ribozymes can be embedded in 5’ or 3’ untranslated region (UTR) of exogenous mRNA expressed in the cell via a viral vector. The presence or the absence of predefined RNA(s) in the cell can induce or suppress the self-cleavage of the ribozyme embedded in the UTR of the exogenous mRNA. As a result, the lifetime of the exogenous mRNA will be reduced or prolonged. That will switch OFF, or turn ON the expression of the exogenous protein encoded into the mRNA. Therefore, oligonucleotide-sensing ribozymes can be used as synthetic riboswitches in the cell in vitro or in vivo for exogenous control of gene expression. The synthetic riboswitches can be designed to sense the presence or absence of disease indicative RNA(s) in the cell and to control the expression of key exogenous or endogenous proteins. Such approach can alter the cell fate, for instance, by inducing apoptosis in the sick cells

only. The approach is universal since the expression of any exogenous protein can be controlled by the presence or absence of any significantly expressed RNA(s) in the cell. Different cancer forms have different RNA expression profiles. Therefore, they have to be treated specifically. However due to its steady proliferation, some cancer cells over-expressed common RNAs. For instance, mRNA of telomerase is over-expressed in many tumor cells. Such RNAs can be used as common effectors for the proposed anticancer treatment in tissues with few normally dividing cells. The synthetic riboswitches reassemble the function of the natural ones found to control gene expression in bacteria, plants, and fungus (Barrick & Breaker, 2007; Winkler & Breaker, 2005). Natural riboswitches are structured RNA domains usually residing in the 5’ UTR of mRNAs. They control gene expression by directly binding a specific metabolite. Bacterial riboswitches execute control of gene expression via self-cleavage, prevention of translation, or termination of transcription. Riboswitches in plants and fungus are found to control gene expression via regulation of alternative splicing (Barrick & Breaker, 2007). Bacterial riboswitches are considered to be novel and very promising targets for antibacterial drug discovery (R. R. Breaker, 2009). The idea of applying ribozymes as a medicine is an old one. Ribozymes have been seen as a tool for gene therapy for a long time (Britton, Larsson, & Ahrlund-Richter, 1994; Freelove & Zheng, 2002; Shaw et al., 2001). The allosteric ribozymes discussed here have the potential to serve as synthetic riboswitches in a human cell, since they have an advance design as discussed in the next sections. They may be used to develop novel therapeutic strategies, which can address many forms of cancer associated with aberrant expression of disease indicative RNA(s). That can be achieved by re-programming the fate of malignant cells through over-expression of exogenous p53 or other key cell factors under the

73


control of synthetic riboswitches. In addition, allosteric ribozymes can be designed to produce conditionally short hairpin (sh) RNAs in the cell inducing microRNAs or RNA interfering(RNAi) pathways (Li, Li, & Rossi, 2006). Such approaches can be used for conditional down regulation of key cellular proteins. These new synthetic biology approaches have the potential to be employed as novel potent therapeutics as described below.

Identifying Cancer Indicative RNAs There are some frequently occurring forms of cancer that are associated with over-expression of certain mRNAs such as mantle cell lymphoma (MCL) and hepatocellular carcinoma (HCC). Lymphoma is a sort of cancer that starts in a lymphocyte, usually in a lymph node or occasionally in other organs. It is divided into two main classes known as Hodgkin lymphoma (HL) and non-Hodgkin lymphoma (NHL). MCL is one of the most severe subtypes of NHL. There are circa sixty thousands new cases of NHL in the United States alone every year. MCL patients make about six percent of all cases of NHL per annum, which means about three and a half thousands new patients in the United States alone. MCL is associated with a genetic mutation that involves chromosomes 11 and 14, called reciprocal translocation t(11;14) in about 85% of patients (Salaverria, Perez-Galan, Colomer, & Campo, 2006). This mutation makes the transfected B lymphocyte (lymphoma cell) to overproduce a cyclin D1 protein. It is known that the cyclin D1 protein directs cell division and growth (Deboer, Vankrieken, Kluinnelemans, Kluin, & Schuuring, 1995). The resulting accumulation of mutant B lymphocytes leads to the formation of MCL tumors. In a very small number of patients t(11;14) mutation is not present. In most of those patients, other genetic mutations cause excess production of cyclin D1. Rarely, MCL can arise from over-expression of other cyclin genes such as cyclin D2 and cyclin D3. Note that the overproduction of cyclin D1

74

protein is due to the over-transcription of cyclin D1 mRNA that has been used as a clinical marker for MCL development (Deboer, et al., 1995). Although the notable progress was achieved in the treatment of MCL with an almost doubling of overall survival over the last three decades, relapses are still very common (Witzig, 2005). Most patients are observed to respond well to initial chemotherapy. Recently, chemotherapy is often combined with stem cell transplantation for patients with MCL. Unfortunately, for most patients, the disease eventually progresses or relapses, despite the therapies employed. In addition, resistance to the chemotherapy frequently develops. Because all these factors, the average progression-free period for patients with MCL is 20 months and the median overall survival prognosis is about four years. Therefore, many researchers continue to look for novel therapies that will prolong remissions and extend survival in patients with MCL, and will have fewer side effects (Goy, 2006). Liver cancer is also a widely spread disorder as the rates of hepatocellular carcinoma (HCC) have been rising by more than 70% in the last 3 decades in the United States alone. Late diagnosis, due to the lack of clinical symptoms, and decreased hepatic function, caused by underlying hepatic disease, have led to the extremely high mortality rates inflicted by HCC (Graveel, et al., 2003). It has been found that over-expression of CRG-L2 mRNA (Cancer related gene-Liver 2) is associated with HCC (Graveel, et al., 2003). The development of novel gene therapy approaches for curing HCC is the main goal of many clinical trials (Hernandez-Alcoceba, et al., 2007). Despite the progress achieved in the life prolongation of patients with terminal phase of HCC, there is still nocure for such patients (Hernandez-Alcoceba, et al., 2007). Normal and malignant cells from a same tissue are isolated from one cancer patient. Total RNA extracts are obtained separately from the isolated normal and malignant cells. The RNA pools are


converted into cDNA molecules by reverse transcription. The DNA molecules can be sequenced using next generation sequencing machines like Solid System IV from Applied Biosystems. Apart from this, the cDNA pools can be fluorescently labeled during the reverse transcription and analyzed on DNA hybridization microarrays such as GeneAtlas System from Affymetrix. The obtained experimental results can be analyzed on parallel computing platforms such as Blade Servers from Dell. These methods can be applied for identifying cancer indicative RNAs in many different disorders as a first step in a comprehensive anticancer therapy as described in the next sections. All these findings strongly suggest that an aberrant expression of normal mRNAs, non-coding RNAs, and the presence of mutant RNAs in tumors can be widely used as reliable molecular markers for diagnostic of many different cancer disorders. Applications of high-throughput RNA expression microarrays and ultra-throughput NGS technologies will certainly increase the number of cancer indicative RNAs known to us. A general scheme for discovery of aberrant expression of RNAs in cancer disorders using NGS or RNA microarray technologies is depicted in Figure 1. Total RNA extracts from one patient can be isolated separately from normal and tumor cells. The extracted RNAs from normal and malignant cells can be transcribed into cDNA libraries using the enzyme reverse transcriptase (RT). NGS machines, such as the Solid System IV from Applied Biosystems or Genome Analyzer from Illuminer, can be employed for ultra-throughput DNA sequencing (Figure 1). Apart from this, DNA hybridization microarrays, such as GeneAtlas System from Affymetrix, can be used for the same goal (Figure 1). In this case, the cDNA molecules have to be fluorescently labeled during the reverse transcription. Both approaches are equally accurate. High-performing computer platforms, such as Blade servers from Dell, can be employed for data manning of the experimental results obtained (Figure 1). Various bioinformatical methods have to be employed

to sort out the primary mutations that cause the cancer from many more secondary mutations during cancer progression. The discovery of cancer indicative RNAs can be of significant importance not only for early molecular diagnostics but also for developing novel comprehensive anticancer therapies. General approaches for anticancer and antiviral therapies based on computationally designed allosteric ribozymes will be discussed in the next sections.

Computational Design of HighSpeed Oligonucleotide-Sensing Allosteric Ribozymes with NOT or YES Logic Functions The hammerhead ribozymes are naturally occurring RNA molecules that possess an autocatalytic function to cleave itself. They are named hammerhead ribozymes because their conserved secondary structure that reassembles the form of a carpenter’s hammerhead (Figure 2). The hammerhead ribozyme catalyzes a transesterification reaction in which the 3’, 5’-phosphodiester is cleaved (Figure 3A and 3B). The reaction yields a cyclic 2’,3’- phosphodiester on nucleotide 17 and a free 5’-hydroxyl on nucleotide 1.1 (Figure 3C). The chemical reaction requires an in-line conformation and divalent cations such as Mg2+ to have an in-line attack from 2’-hydroxyl of the nucleotide 17 to the scissile phosphate. Therefore, the free 2’-OH on nucleotide 17 and divalent cations are essential for the catalytic function of the hammerhead ribozyme (Figure 2). The hammerhead ribozyme in nature is involved in the replication of viroids and some satellite RNAs. It is found both in animals and plants. Note that the conserved secondary structure of the hammerhead ribozyme defines a certain conserved tertiary (3D) structure that is essential for the catalytic function of the ribozyme (Martick & Scott, 2006). There are two main variants of the hammerhead ribozyme, which differ in their

75


Figure 1. General strategies for identifying cancer indicative RNAs in humans applying either ultrathroughput next generation sequencing technology or DNA hybridization microarrays

kinetics and concentration of divalent cations needed for the catalytic function. One variant of the hammerhead ribozyme is called minimal (Figure 2A) and the other is called extended (Figure 2B). As suggested by its name the minimal variant of the hammerhead ribozyme is the shortest possible sequence of the ribozyme that has all essential structures including stem I, II, and III, and the catalytic core (Figure 2A). In contrast, the extended variant of the ribozyme has an additional bulge loop in stem I (Figure 2B). The nucleotides in this loop make contacts with the nucleotides from the loop in stem II in the extended hammerhead ribozyme (Figure 2B). It is found that stems I and II have to get close together in the 3D conformation of the hammerhead ribozyme for its efficient catalysis. It is believed that divalent cations (such as Mg2+) bridge stems I and II. Because of its tertiary contacts between stems I and II, the extended variant of the hammerhead ribozyme requires only 1 mM Mg2+ to reach its maximal

76

speed of cleavage. In contrast, the minimal variant of the hammerhead ribozyme requires 10 mM Mg2+ to reach its maximal speed of cleavage. In addition, maximal speed of the extended variant of the hammerhead ribozyme is 10 fold higher than that of the minimal variant. Note that the physiological concentration of Mg2+ in the cell is 1 mM. That makes the extended variant of the hammerhead ribozyme (Figure 2B) suitable to serve as a pattern construct for designer biosensors that have a cellular function. Therefore, the extended hammerhead ribozyme from the human parasite Schistosoma mansoni is used as a pattern construct for all designer hammerhead ribozymes that are intended to have a cellular function. The wild-type schistosoma hammerhead ribozyme possesses high-speed kinetics of cleavage that can reach a speed of circa 10 per min. Moreover, schistosoma hammerhead ribozyme has been employed as a regulatory element for exogenous control of gene expression in human cell line (HEK293T) as well as in mouse.


Figure 2. Three variants of the hammerhead ribozyme

However, these experiments are not very useful since the wild-type schistosoma hammerhead ribozyme is not an allosteric molecule and cannot sense any specific effector in the cell apart from general RNA inhibitors that are toxic for the whole cell (Yen, et al., 2004). Many promising applications of computationally designed allosteric ribozymes that can sense

specific RNA or small molecules are described in the next sections. The allosteric domain for all designer ribozymes is introduced in stem III (Figure 2C) to preserve all tertiary contacts between stem I and II existing in the wild-type schistosoma hammerhead ribozyme. This maintains the high-speed kinetics of the designer ribozymes. The allosteric domain can sense the presence of either a RNA (or DNA) sequence (Figures 4 and 6) or a small molecule (Figure 5). If the effector is a small molecule, the allosteric domain contains a RNA aptamer (Figure 5). Aptamers are RNA or DNA molecules that specifically bind a predefined molecule. They are obtained by a procedure called in vitro selection of nucleic acids. Up to now, there are only two synthetic RNA aptamers, which are found to be able to work in the cells. These aptamers sense theophylline (Figure 5A and 5B) or tetracycline (Figure 5C). The natural riboswitches also have aptamer domains that can bind a metabolite with a great affinity and specificity. For the oligonucleotide-sensing ribozyme, the allosteric domain is complementary to the effector molecule (Figures 4 and 6). All ribozymes are designed by a published and patented computational procedure that has two main parts. In the first part, the algorithm computes two different states (conformations) of an allosteric ribozyme in the presence or absence of effector. The different conformations in the presence or absence of effector are computed for their thermodynamic stability and kinetics of folding. The thermodynamic stability is computed by a partition function for RNA folding that gives to all possible RNA secondary structures a probability for a formation based on their thermodynamic stability. This produces very accurate designer ribozymes (Penchovsky & Breaker, 2005). One of the states has an active conformation, in which all three stems I, II, and III are formed as in the wild-type schistosoma hammerhead ribozyme (Figures 2C, 4A, 5A, and 6B). In an active conformation, the ribozyme cleaves its substrate.

77


Figure 3. The catalytic mechanism of the hammerhead ribozyme

Figure 4. A high-speed oligonucleotide-sensing ribozyme construct with NOT Boolean logic function

In an inactive conformation, the ribozyme forms stem IV instead of stem III (Figures 4B, 5B, and 6A). As a result, the catalytic core of the hammerhead ribozyme is not formed and the allosteric molecule cannot cleave itself or its target molecule. In a case of an allosteric ribozyme with a NOT logic function, the active state is without the effec-

78

tor while the ribozyme is inactive in the presence of effector (Figures 4A and 5A).In contrast, an allosteric ribozyme with YES Boolean logic function is active in the presence of effector (Figure 6B) and inactive in its absence (Figure 6A). Note that the accuracy of the algorithm does not suffer from any non-computed tertiary interactions because there are kept constant for both computed


Figure 5. A high-speed small molecule-sensing ribozyme with NOT Boolean logic function

Figure 6. A high-speed oligonucleotide-sensing ribozyme with YES Boolean logic function

states. The properties of the allosteric ribozyme can be tuned by changing the thermodynamic gap between the ON and the OFF states. The applied algorithm exhibits more than 90% accuracy (Penchovsky & Breaker, 2005). It can produce RNA sequences for several allosteric ribozymes in a timescale of several min. All ribozymes are designed in a modular form so that the specificity

of the allosteric domain can be easily altered applying reverse folding, which is implemented in the second part of the computational procedure. The computational procedure discussed here is more efficient than any in vitro selection or rational design method employed for production of allosteric ribozymes. It avoids many limitations of in vitro selection procedures such as a limited

79


initial pool, arising “mini monster” molecules that replicate faster than all other molecules, evolving of opportunistic strategies for bypassing of the selection step, and the long time needed for performing all biochemical procedures in many cycles. Rational design methods do not explore many possible variants, since they are not computerized. The computational procedure discussed here is very time-efficient because it can test hundred thousand different sequences for a few hours on an average personal computer employing a random search algorithm. In addition, this computational procedure avoids all handicaps of in vitro selection and rational design methods accurately producing allosteric ribozymes with desired potteries as experimentally verified (Penchovsky & Breaker, 2005). The experimental results obtained from the designer ribozymes demonstrate a huge dynamic range (over 1000 fold) in kinetics of cleavage between the presence and absence of effector (Penchovsky & Breaker, 2005). The allosteric ribozymes show high-speed kinetics cleaving to 80% within one min under physiological relevant conditions such as 1 mM MgCl2, 90 mM KCl, 25 mM NaCl, 50 mM Tris-HCl (pH 7.5 at 37ºC) (Penchovsky & Breaker, 2005). The designer ribozymes possess high-specificity of activation by their effector molecules because they can distinguish two mismatching bases out of a 22 nt long oligomer. All this makes the designer ribozymes very suitable as synthetic riboswitches in vivo for various gene expression control applications as discussed in the next sections.

Allosteric Ribozymes that Function as Synthetic Riboswitches for Exogenous Control of Gene Expression and Reprogramming the Cell Fate Applications of computationally designed allosteric ribozymes as synthetic riboswitches can open new avenues to the development of universal

80

strategies for exogenous control of gene expression and re-programming the cell fate. These novel molecular bioengineering methods may have many practical applications to nanobiotechnology and pharmaceutical industry in the near future. At the first place, the oligonucleotide-sensing allosteric ribozymes can be employed as synthetic riboswitches that control gene expression of key exogenous proteins. The approach can be used as a universal tool for reprogramming the cell fate by inducing apoptosis through conditional expression of exogenous proteins such as p53, PUMA, and many others. This general approach may have many far-reaching applications to the pharmaceutical and biotech industries including novel gene therapies of cancer. A construct containing a p53 gene with a strong promotor under the control of RNA-sensing allosteric ribozymes can be delivered into a tumor tissue by a viral expression vector (Figure 7A). The viral vector should be chosen in regard to the tissue. In the common case, the viral vector should be non-integrating such as adenovirus, adeno-associated virus, herpes virus, and others. All cells, regardless tumor or normal, must be infected by the viral vector (Figure 7B). However, due to the function of the RNA-sensing ribozymes the exogenous p53 gene will be expressed in the tumor cells only because they carry cancer indicative(ci) RNAs (Figure 7C). Normal cells do not have ciRNAs and will not be able to express the exogenous p53 (Figure 7C). As a result, in the tumor cells the apoptosis is going to be induced, and they are going to die (Figure 7C). In contrast, the normal cells are going to survive the treatment (Figure 7C). This is all about the basic idea for curing any type of cancer by finding tools to kill all tumor cells while keeping alive most of the healthy cells. The implementation of this general idea with RNA-sensing allosteric ribozymes that work as synthetic riboswitches is explained in details below. Let us assume that the pathological phenotype of a tumor cell is defined by the co-expression of


Figure 7. A general scheme for anticancer gene therapy based on RNA-sensing allosteric ribozymes

two ci mRNAs. A general anticancer gene therapy, depicted in Figures 8 and 9, can be employed to induce the apoptosis in all cancer cells while sparing the normal ones. The same scheme can be easily simplified for a single ciRNA sequence when the allosteric domains of both ribozymes are identical. A synthetic gene construct expresses a synthetic mRNA molecule that contains exogenous p53 and two allosteric ribozymes with a NOT logic function. The ribozymes are designed to sense the presence of two ciRNAs (Figure 8 and 9). The construct is delivered into the cell via a viral vector. If the cell is normal the two ciRNAs are not going to be expressed. As a result, both

NOT riboswitches are going to fold into an active conformation in which all three stems I, II, and II are properly formed (Figure 8A). Therefore, the ribozymes will cleave themselves at both ends (Figure 8B). The exogenous mRNA will decay rapidly due to the absence of a cap structure at the 5’-end and a polyA tail at the 3’-end (Figure 8B). Thus, the exogenous p53 gene is not going to be expressed and a healthy cell will survive the treatment (Figure 8C). The opposite events will be happening in the tumor cells where both ciRNAs are present (Figure 9). The ciRNAs are going to hybridize with the allosteric domains of the ribozymes (Figure 9A). That will result in

81


Figure 8. Two synthetic riboswitches with a NOT logic function suppress the expression of exogenous p53 in the absence of tumor indicative RNAs

formation of stem IV instead of stem III inactivating the catalytic function of the ribozymes (Figure 9B). The lifetime of the exogenous mRNA will be close to normal. The synthetic p53 gene will be spliced (Figure 9C) and will be over-expressed. All that will result in induction of apoptosis in the tumor cells and their death (Figure 9D). Note that the allosteric ribozymes can perfectly well distinguish effector oligonucleotides with only two mismatches out of 22 nt. That is sufficient to ensure ribozyme activation only in the presence of a specific effector oligonucleotide avoiding any unspecific activation by other RNAs in the cell. The described anticancer therapy can be potentially applied to many cancer types and other disorders.

82

The p53 is not the only protein that has to be over-expressed to induce apoptosis. For instance, it is observed selection against PUMA gene expression in Myc-driven B-Cell lymphoma genesis. This means that exogenous over-expression of PUMA in these cells may be more suitable than that of p53 (Garrison et al., 2008). Moreover, selective over-expression of exogenous Arf in some lymphoma cells can lead to suppression of Myc-driven lymphoma genesis (Bouchard et al., 2007). The proposed therapy is universal since the ribozymes can be designed to sense the presence of any sufficiently expressed RNA(s) and can control the expression of any exogenous gene. This makes the method very adaptive in address-


Figure 9. Two synthetic riboswitches with a NOT logic function induce the expression of exogenous p53 in the presence of tumor indicative RNAs

ing various disorders. Thus, the proposed strategy for exogenous control of gene expression can be applied in many regulatory circuits in vivo. For instance, one of the promising real-world applications would be addressing some forms of cancer such as HCC and MCL because we can transfect liver and B-cells in vivo with high efficiency maintaining constant viral expression for at least two weeks. In addition, there are ciRNAs that are over-expressed during the early stage of HCC and MCL development as detailed in the introduction. In some types of cancer and other disorders, specific ciRNAs are no expressed. This is the op-

posite case in which health indicative(hi) RNAs are suppressed. In the general case, cell growth suppressor genes can be inhibited in some tumors. In such cases, a p53 gene (or another synthetic gene) has to be expressed only in the absence of hiRNAs. For this aim, riboswitches with a YES Boolean logic function (Figure 6) have to be employed instead of NOT gates (Figure 4) as depicted in Figures 10 and 11. A synthetic p53 gene under the control of two riboswitches with a YES logic function is delivered into the cell via a viral vector (Figure 7). If the transfected cell is normal hiRNAs are expressed (Figure 10). The hiRNAs hybridize with allosteric domains of the ribozymes. As a result, the YES gates are turned on (Figure 10A). They undergo a self-cleavage (Figure 10B). Therefore, the synthetic mRNA is rapidly decayed and the exogenous p53 is not expressed. The healthy cell survives the treatment. In contrast, the tumor cell does not have the two hiRNAs (Figure 11). As a result, both riboswitches are inactive since stem IV is formed instead of stem III (Figure 11A). Therefore, the exogenous mRNA has a normal lifetime. The p53 mRNA is spliced (Figure 11B) and translated into the exogenous p53 protein. As a result, the tumor cell undergoes apoptosis and dies. Both methods for exogenous control of gene expression described above can have countless concrete applications to many different disorders, since they are truly universal in terms of effector RNAs and exogenous proteins controlled. The general approaches described in this section have the potential to be adopted for addressing many human disorders that have no efficient treatment up to date. The designer ribozymes discussed here offer new efficient tools for re-programming the cell fate. They are easily to design and to test experimentally and can be widely employed as cellular biosensors for building gene expression control circuits in any species of interest that can be transfected by an expression vector. Small molecule-sensing allosteric ribozymes (Figure 5) can be also employed as synthetic ribo-

83


Figure 10. Synthetic riboswitches with a YES logic function suppresses the expression of exogenous p53 in the presence of health indicative RNAs

switches for exogenous control of gene expression as the same methods can be applied like in the cases with oligonucleotide-sensing ribozymes (Figures 8-11). Small molecule-sensing allosteric ribozymes may have useful applications in stem cell research. Instead of controlling expression of exogenous p53, cell growth and differentiation protein factors can be put under the control of synthetic riboswitches. Stem cell differentiation can be triggered by a drug that specifically binds

84

to a synthetic riboswitch in the cell and turn on the expression of key proteins. Another wide-spread application of small molecule-sensing ribozymes would be for tackling certain chronic diseases, which require a long term recapitulative treatment. For instance, some diabetic patients need to take insulin when their blood sugar’s level rises too high. At the same time, they have to be very careful not to take too much insulin because the low level of the blood sugar is a life threatening condition. Therefore,


Figure 11. Synthetic riboswitches with a YES logic function induce the expression of exogenous p53 in the absence of health indicative RNAs

an exogenous insulin gene can be put in the cell under the control of a glucose-sensing synthetic riboswitch with a YES Boolean logic function like shown in Figures 10 and 11. This approach, however, is technically much more complicated than the cases with the oligonucleotide-sensing ribozymes. The main reason for the complication concerns an additional step needed for in vitro

selection of a glucose-binding RNA aptamer that has to be successfully accomplished previous to the design of a glucose-sensing ribozyme. In vitro selection methods have been proven to be efficient for obtaining many RNA aptamers that selectively bind different small molecules. Therefore, despite the additional efforts needed for the in vitro selection of a glucose-binding

85


RNA aptamers that research maybe very helpful for many diabetic patients worldwide.

Conditional Production of ShortHairpin RNAs by Allosteric Ribozymes with YES Logic Function for Gene Down Regulation through RNA Interfering and microRNA Pathways The discovery of RNA interfering (RNAi) and microRNA pathways in mammalians including humans raises many hopes for the deployment of novel and efficient tools for gene therapy of viral infections and other wide-spread diseases such as cancer. That can be achieved by specific targeting of key for the disease’ development mRNAs or viral RNAs through microRNA and RNA silencing pathways (Dorsett & Tuschl, 2004). Despite the results achieved in the expression of shRNAs in mammalian cells and the RNA silencing effect demonstrated, there are many side effects that spell significant problems for the application of RNAi technology in the pharmaceutical industry. A major problem for the efficient retroviral RNAs silencing, including HIV, is the fact that viral RNAs tend to mutate very fast under the selection pressure of constantly expressed interfering RNAs. As a result, after certain viral generation the specific interfering RNAs are not anymore effective against the virus RNA. For constantly expressed interfering RNAs strong silencing effect of many non-target mRNAs is often observed. To avoid these problems, interfering RNAs like shRNAs should be temporarily expressed only when needed. That can be achieved by applying RNA-sensing allosteric ribozymes with a YES logic function as depicted in Figure 12. An allosteric ribozyme with YES logic function can be constantly expressed into the cell using viral expression vectors. Without effector RNAs the YES gate (Figure 4) folds into the inactive state in which stem IV is formed instead of stem III (Figure 12A). In this conformation, the ribozyme’s self-cleavage is inhibited. An effector RNA strand 86

with a predefined sequence can bind to the ribozyme’s allosteric domain and can activate the ribozyme by making stem III instead of stem IV (Figure 4 & 12B). As a result, the ribozyme’s self-cleavage can produce a short-hairpin(sh) RNA Figure 12C. The shRNA can be converted in the cell into double-stranded(ds) RNA with sticky ends by the Dicer enzyme Figure 12D. The dsRNA will be made single-stranded by the RISC complex. The ssRNA in the RISC complex will be used as a template to degrade targeted mRNA or for microRNA to translational repression (Figure 12E). When the effector RNAs, that can be viral and disease indicative RNAs, are not present anymore the ribozyme will form inactive state again (Figure 12A) and shRNA production will cease. The restricted time of shRNA production will not allow the targeted viral RNA to mutate under the selection pressure of the shRNA and to escape the silencing effect. In addition, it will be avoided long term non-specific silencing for many mRNAs. Therefore, the conditional expression of shRNA can reduce the side-effects of RNA interfering technology to a minimum making it more feasible to tackle various disorders such as viral infection diseases including AIDS.

CONCLUSION The procedure for fully computerized design of allosteric ribozymes offers completely new opportunities for an efficient production of highprecision biosensors with many in vitro and in vivo applications to nanobiotechnology, pharmaceutical and biotech industries. Perhaps the most promising application of the computationally designed allosteric ribozymes is for exogenous control of gene expression either in cell lines or in vivo. The ribozymes can be designed to sense the presence or the absence of any significantly expressed RNA in the cell. In addition, they can be engineered to sense the presence of small molecules. Allosteric ribozymes can be delivered in the cell with a viral vector. They can be used


Figure 12. A scheme for producing short-hairpin (sh) RNAs for siRNA and microRNA gene targeting by allosteric ribozymes with a YES Boolean logic function

as synthetic riboswitches for the control of gene expression of any exogenous protein. This approach can be applied for re-programming the cell fate by over-expression of key regulatory proteins such as p53, PUMA, and many others. This approach for re-programming the cell fate may be used as a comprehensive therapy of many human disorders including many forms of cancer. There are only two preconditions for the execution of the proposed therapy. The first precondition concerns the necessity for the presence of cancer indicative RNA(s). There are many forms of cancer such as MCL and HCC that are clearly associated with over-expression of certain RNAs. Moreover, application of HGS and DNA microarray technologies enables researchers to discover easily new RNAs that are indicative of a

particular type of cancer. The second precondition concerns the existence of viral vectors that can efficiently deliver in the cell a exogenous gene under the control of synthetic riboswitches. There are many human viral vectors such as herpes vectors that can transfect human cells in vivo with highefficiency. At the same time, they are not toxic for humans since all pathogenic viral genes are being excluded from the viral sequence. The field of gene therapy is rapidly improving and offering new more efficient and safe human viral vectors despite some setbacks in the past. If both preconditions are fulfilled for a particular type of cancer, the synthetic riboswitches can be employed as a cure after preclinical and clinical trails. Note the difference between the method proposed here and many already tested methods

87


in various clinical trails based on over-expression of exogenous p53 in tumor cells. In all previous methods, the exogenous p53 gene was delivered and over-expressed in any cell of the targeted tissue regardless, cancerous or normal (Lang et al., 2003; Rakozy et al., 1999; Weber & Wenz, 2002). As a result, apoptosis was induced not only in tumor but also in normal cells. To reduce the toxicity for the normal cells the approach was modified by applying viral vectors that transfect tumor cells with higher efficiency than normal cells. However, these approaches are found to be not good enough at distinguishing between tumor and normal cells. When the viral titer was high enough to transfect most of the tumor cells many normal cells were found to be also transfected. Unfortunately, when the viral titer was low enough to transfect a few normal cells; it was not sufficient to transfect most of the tumor cells. The obvious defect of using exogenous p53 for anticancer gene therapy is resolved by the proposed method based on RNA-sensing synthetic riboswitches. A high-viral titer can be used for transfecting all tumor cells as well as few normal cells. This will not be toxic for the normal cells since exogenous p53 will be over-expressed mainly if not only in the tumor cells. Allosteric ribozymes can be also designed to produce conditionally shRNAs in the cell inducing RNAi and microRNA pathways. This can be used as an antiviral tool. Besides, the approach can be employed for conditional down regulation of any endogenous gene. In general, the proposed methods can be employed as universal tools for re-programming the cell fate both in vitro and in vivo. There are still experiments in vivo including clinical trails to be done to cure a particular disorder in humans. However, all essential steps in a comprehensive drug development strategy are clearly laid out and the chances for new efficient drug development must be taken especially in the cases without any cure available.

88

Small molecule-sensing ribozymes can be also used for exogenous control of gene expression in a similar way like oligonucleotide-sensing ribozyme (Win & Smolke, 2008). The exogenous gene can be induced by small molecule drugs such as theophylline and tetracycline. This approach, however, has to overcome more hurdles than that of the RNA-sensing ribozyme. The small moleculesensing ribozymes have a much smaller dynamic range of few hundred fold (Link et al., 2007) than that of the RNA-sensing ribozymes with a dynamic range of several thousand fold (Penchovsky & Breaker, 2005) in their activation. This makes the exogenous control of gene expression by RNAsensing ribozyme much more robust than that of small molecule-sensing ribozymes. This problem may be overcame in some degree by using the computational design procedure for producing small molecule-sensing ribozyme like this for oligonucleotide-sensing ribozymes (Penchovsky & Breaker, 2005). In addition, there are only two aptamers, for theophylline and tetracycline, can be used in vivo. Both small molecule drugs have a lot of side effects. In contrast, oligonucleotidesensing ribozyme can be designed to sense the presence of any RNA sequence, which makes them universal biosensors with a wide range of applications. However, small molecule-sensing ribozymes can be very well suited for stem cell applications. Stem cells can be made to start certain differentiation after exposure to a drug using small molecule-sensing synthetic riboswitches. Such approach may advance the field of stem cell therapy and may have many practical applications to various stem cell therapies. Apart from being used as synthetic riboswitches in the cell, the ribozymes presented have many applications as biosensors in vitro, which will be discussed elsewhere. Engineering gene regulatory circuits in the cell based on computationally designed allosteric ribozymes can have a profound impact to the fields of applied synthetic biology (Isaacs, Dwyer, & Collins, 2006), drug devolve-


ment (R. R. Breaker, 2005; Eisenstein, 2005), and nanobiotechnology (Margolin & M.N., 2005).

ACKNOWLEDGMENT I thank Prof. Hans Lehrach from Max Planck Institute for Molecular Genetics in Berlin, Germany for helpful discussions and Dimitar Todorov, from Sofia University “St. Kliment Ohridski”, for critically reading the manuscript. Robert Penchovsky’s research is supported by a grant (DDVU02/5/2010) awarded by the Bulgarian National Science Fund (BNSF).

REFERENCES Barrick, J. E., & Breaker, R. R. (2007). The power of riboswitches. Scientific American, 296(1), 50–57. doi:10.1038/scientificamerican0107-50 Bateman, A., & Quackenbush, J. (2009). Bioinformatics for next generation sequencing. Bioinformatics (Oxford, England), 25(4), 429. doi:10.1093/bioinformatics/btp037 Bentley, G., Higuchi, R., Hoglund, B., Goodridge, D., Sayer, D., & Trachtenberg, E. A. (2009). High-resolution, high-throughput HLA genotyping by next-generation sequencing. Tissue Antigens, 74(5), 393–403. doi:10.1111/j.13990039.2009.01345.x Blount, K., Puskarz, I., Penchovsky, R., & Breaker, R. (2006). Development and application of a high-throughput assay for glmS riboswitch activators. RNA Biology, 3(2), 77–81. doi:10.4161/ rna.3.2.3102 Bouchard, C., Lee, S., Paulus-Hock, V., Loddenkemper, C., Eilers, M., & Schmitt, C. A. (2007). FoxO transcription factors suppress Myc-driven lymphomagenesis via direct activation of Arf. Genes & Development, 21(21), 2775–2787. doi:10.1101/gad.453107

Breaker, R., & Penchovsky, R. (2008). Computational design of ribozymes. Patent Application, WO-2008/127382-A2, 81. Breaker, R. R. (2005). Gene expression control: Harnessing RNA switches. Gene Therapy, 12(9), 725–726. doi:10.1038/sj.gt.3302461 Breaker, R. R. (2009). Riboswitches: From ancient gene-control systems to modern drug targets. Future Microbiology, 4, 771–773. doi:10.2217/ fmb.09.46 Britton, S., Larsson, S., & Ahrlund-Richter, L. (1994). Ribozyme-based gene therapy of HIV. Lakartidningen, 91(39), 3491–3494. Chan, E. Y. (2009). Next-generation sequencing methods: Impact of sequencing accuracy on SNP discovery. Methods in Molecular Biology (Clifton, N.J.), 578, 95–111. doi:10.1007/978-160327-411-1_5 Deboer, C. J., Vankrieken, J. H. J. M., Kluinnelemans, H. C., Kluin, P. M., & Schuuring, E. (1995). Cyclin D1 messenger-Rna overexpression as a marker for mantle cell lymphoma. Oncogene, 10(9), 1833–1840. Dorsett, Y., & Tuschl, T. (2004). siRNAs: Applications in functional genomics and potential as therapeutics. Nature Reviews. Drug Discovery, 3(4), 318–329. doi:10.1038/nrd1345 Eisenstein, M. (2005). Computational evolution offers riboswitch solution. Nature Methods, 2(12), 894. doi:10.1038/nmeth1205-894 Freelove, A. C., & Zheng, R. (2002). The power of ribozyme technologies: The logical way ahead for molecular medicine and gene therapy? Current Opinion in Molecular Therapeutics, 4(5), 419–422.

89


Garrison, S. P., Jeffers, J. R., Yang, C. Y., Nilsson, J. A., Hall, M. A., & Rehg, J. E. (2008). Selection against PUMA gene expression in Myc-driven B-cell lymphomagenesis. Molecular and Cellular Biology, 28(17), 5391–5402. doi:10.1128/ MCB.00907-07

Li, M., Li, H., & Rossi, J. J. (2006). RNAi in combination with a ribozyme and TAR decoy for treatment of HIV infection in hematopoietic cell gene therapy. Annals of the New York Academy of Sciences, 1082, 172–179. doi:10.1196/ annals.1348.006

Goy, A. (2006). New directions in the treatment of mantle cell lymphoma: An overview. Clinical Lymphoma & Myeloma, 7, S24–S32. doi:10.3816/ CLM.2006.s.005

Liang, S., Lu, Y., Jelinek, J., Estecio, M., Li, H., & Issa, J. P. (2009). Analysis of epigenetic modifications by next generation sequencing. Conference Proceedings;... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Conference, 2009, 6730.

Graveel, C. R., Harkins-Perry, S. R., Acevedo, L. G., & Farnham, P. J. (2003). Identification and characterization of CRG-L2, a new marker for liver tumor development. Oncogene, 22(11), 1730–1736. doi:10.1038/sj.onc.1206309 Guryev, V., & Cuppen, E. (2009). Next-generation sequencing approaches in genetic rodent model systems to study functional effects of human genetic variation. FEBS Letters, 583(11), 1668–1673. doi:10.1016/j.febslet.2009.04.020 Hernandez-Alcoceba, R., Sangro, B., & Prieto, J. (2007). Gene therapy of liver cancer. Annals of Hepatology, 6(1), 5–14. Isaacs, F. J., Dwyer, D. J., & Collins, J. J. (2006). RNA synthetic biology. Nature Biotechnology, 24(5), 545–554. doi:10.1038/nbt1208 Jung, C. H., Hansen, M. A., Makunin, I. V., Korbie, D. J., & Mattick, J. S. (2010). Identification of novel non-coding RNAs using profiles of short sequence reads from next generation sequencing data. BMC Genomics, 11, 77. doi:10.1186/14712164-11-77 Lang, F. F., Bruner, J. M., Fuller, G. N., Aldape, K., Prados, M. D., & Chang, S. (2003). Phase I trial of adenovirus-mediated p53 gene therapy for recurrent glioma: Biological and clinical results. Journal of Clinical Oncology, 21(13), 2508–2518. doi:10.1200/JCO.2003.11.138

90

Link, K. H., Gu, L. X., Ames, T. D., Yen, L. S., Mulligan, R. C., & Breaker, R. R. (2007). Engineering high-speed allosteric hammerhead ribozymes. Biological Chemistry, 388(8), 779–786. doi:10.1515/BC.2007.105 Margolin, A. A., & M.N., S. (2005). Boolean calculation made easy (for ribozyme). Nature Biotechnology, 23, 2. doi:10.1038/nbt1105-1374 Martick, M., Horan, L. H., Noller, H. F., & Scott, W. G. (2008). A discontinuous hammerhead ribozyme embedded in a mammalian messenger RNA. Nature, 454(7206), 899–902. doi:10.1038/ nature07117 Martick, M., & Scott, W. G. (2006). Tertiary contacts distant from the active site prime a ribozyme for catalysis. Cell, 126(2), 309–320. doi:10.1016/j. cell.2006.06.036 Morrissy, A. S., Morin, R. D., Delaney, A., Zeng, T., McDonald, H., & Jones, S. (2009). Next-generation tag sequencing for cancer gene expression profiling. Genome Research, 19(10), 1825–1835. doi:10.1101/gr.094482.109 Muller, U. F. (2009). Evolution of ribozymes in an RNA world. Chemistry & Biology, 16(8), 797–798. doi:10.1016/j.chembiol.2009.08.001


Oetting, W. S. (2010). Impact of next generation sequencing: The 2009 Human Genome Variation Society Scientific Meeting. Human Mutation, 31(4), 500–503. doi:10.1002/humu.21210 Penchovsky, R., Birch-Hirschfeld, E., & McCaskill, J. S. (2000). End-specific covalent photodependent immobilisation of synthetic DNA to paramagnetic beads. Nucleic Acids Research, 28(22), E98. doi:10.1093/nar/28.22.e98 Penchovsky, R., & Breaker, R. R. (2005). Computational design and experimental validation of oligonucleotide-sensing allosteric ribozymes. Nature Biotechnology, 23(11), 1424–1433. doi:10.1038/nbt1155

Shaw, L. C., Skold, A., Wong, F., Petters, R., Hauswirth, W. W., & Lewin, A. S. (2001). An allelespecific hammerhead ribozyme gene therapy for a porcine model of autosomal dominant retinitis pigmentosa. Molecular Vision, 7, 6–13. Tang, J., & Breaker, R. R. (1997). Rational design of allosteric ribozymes. Chemistry & Biology, 4(6), 453–459. doi:10.1016/S1074-5521(97)90197-6 Teng, X., & Xiao, H. (2009). Perspectives of DNA microarray and next-generation DNA sequencing technologies. Science in China. Series C, Life Sciences, 52(1), 7–16. doi:10.1007/s11427-0090012-9

Porta, H., & Lizardi, P. M. (1995). An allosteric hammerhead ribozyme. Biotechnology, 13(2), 161–164. doi:10.1038/nbt0295-161

Weber, K. J., & Wenz, F. (2002). p53, apoptosis and radiosensitivity - Experimental and clinical data. Onkologie, 25(2), 136–141. doi:10.1159/000055223

Rakozy, C., Grignon, D. J., Li, Y., Gheiler, E., Gururajanna, B., & Pontes, J. E. (1999). p53 gene alterations in prostate cancer after radiation failure and their association with clinical outcome: A molecular and immunohistochemical analysis. Pathology, Research and Practice, 195(3), 129–135.

Willenbrock, H., Salomon, J., Sokilde, R., Barken, K. B., Hansen, T. N., & Nielsen, F. C. (2009). Quantitative miRNA expression analysis: Comparing microarrays with next-generation sequencing. RNA (New York, N.Y.), 15(11), 2028–2034. doi:10.1261/rna.1699809

Ramos, A. M., Crooijmans, R. P., Affara, N. A., Amaral, A. J., Archibald, A. L., & Beever, J. E. (2009). Design of a high density SNP genotyping assay in the pig using SNPs identified and characterized by next generation sequencing technology. PLoS ONE, 4(8), e6524. doi:10.1371/ journal.pone.0006524

Win, M. N., & Smolke, C. D. (2008). Higherorder cellular information processing with synthetic RNA devices. Science, 322(5900), 456–460. doi:10.1126/science.1160311

Reis-Filho, J. S. (2009). Next-generation sequencing. Breast Cancer Research, 11(Suppl 3), S12. doi:10.1186/bcr2431 Salaverria, I., Perez-Galan, P., Colomer, D., & Campo, E. (2006). Mantle cell lymphoma: From pathology and molecular pathogenesis to new therapeutic perspectives. Haematologica, 91(1), 11–16.

Winkler, W. C., & Breaker, R. R. (2005). Regulation of bacterial gene expression by riboswitches. Annual Review of Microbiology, 59, 487–517. doi:10.1146/annurev.micro.59.030804.121336 Witzig, T. E. (2005). Current treatment approaches for mantle-cell lymphoma. Journal of Clinical Oncology, 23(26), 6409–6414. doi:10.1200/ JCO.2005.55.017

91


Wyman, S. K., Parkin, R. K., Mitchell, P. S., Fritz, B. R., O’Briant, K., & Godwin, A. K. (2009). Repertoire of microRNAs in epithelial ovarian cancer as determined by next generation sequencing of small RNA cDNA libraries. PLoS ONE, 4(4), e5311. doi:10.1371/journal.pone.0005311 Yen, L., Svendsen, J., Lee, J. S., Gray, J. T., Magnier, M., & Baba, T. (2004). Exogenous control of mammalian gene expression through modulation of RNA self-cleavage. Nature, 431, 5. doi:10.1038/nature02844

KEY TERMS AND DEFINITIONS Allosteric Ribozymes: Are engineered functional ribonucleic acid (RNA) molecules. Their catalytic function is modulated by effector molecules. The allosteric domain is distant from the catalytic center of the ribozyme. Apoptosis: A set of many different biochemical pathways in the cell. When induced, it can lead to cell death. Computational Design of Allosteric Ribozymes: A computerized approach that employs different algorithms to find RNA sequences of allosteric ribozymes with desired properties by testing a large pool of random sequences.

92

Gene Therapy: is a set of methods for curing different diseases by employing viral vectors. Exogenous Control of Gene Expression: An engineering approach for a regulation of synthetic genes. In Vitro Selection: A set of different biochemical methods used for obtaining RNA, DNA, and polypeptides with desired properties by repeatedly applying amplification, mutation, and selection steps over an initial pool with random sequences. Riboswitches: RNA molecules that control gene expression by directly binding effector molecules. Natural riboswitches are discovered in bacterial and plant species. They are usually found in the untranslated regions of mRNAs. Synthetic riboswitches can be designed to control expression of synthetic genes. Ribozymes: are functional RNA molecules that can catalyze specific biochemical reactions such as self-cleavage, self-splicing, and others. They are found in plants and animals. They can accelerate the rate of specific biochemical reactions by 1011 fold. They have conserved secondary and tertiary structures. RNA Aptamers: Are structured molecules that can bind to a specific ligand. RNA aptamers are usually engineered by an in vitro selection procedure. There are naturally occurring aptamers in bacterial and plant species as a part of a riboswitch.

Section 3

Medical Treatment and Research: Ethics and Applications

94

Chapter 6

Ethical Guidelines for the Quality Assessment of Healthcare Amit Chattopadhyay National Institutes of Health (NIDCR), USA Khushdeep Malhotra National Institutes of Health (NIDCR), USA Sharmila Chatterjee George Washington University, USA

ABSTRACT One of the keystone concepts of healthcare improvement is the belief that informed choices will lead to enhanced quality of life and healthcare. Medical Ethics, or Bioethics, is the study of moral issues in the fields of medical treatment and research. It is also used to describe ethical issues in the life sciences and the distribution of scarce medical resources. This chapter will review and describe the principles of ethics and discuss how ethical principles can be used as guidelines for the quality assessment of healthcare provision. It will also discuss areas such as: ethical handling of information, patient safety, communication, obligation for impartial quality assessment, private health information protection, ethical committees and supervision authorities, competence of the assessor, supervision of ethical guidelines for health quality assessment, research and publication ethics, and global ethics of healthcare. Another goal of the manuscript will be to serve as a central reference to access of information about resources related to this topic.

INTRODUCTION As a branch of Philosophy, Ethics guides the leading of a good life. Ethics also guides moral conduct in life. Applied Ethics, on the other hand, is a discipline that guides application of ethical

theory to real-life situations. Medical Ethics or Bioethics, studies the role of value judgments and “moral issues in the fields of medical treatment and research. The term is also sometimes used more generally to describe ethical issues in the life sciences and the distribution of scarce medi-

DOI: 10.4018/978-1-61350-120-7.ch006


Ethical Guidelines for the Quality Assessment of Healthcare

cal resources. The professional fields that deal with ethical issues in medicine include medicine, nursing, law, sociology, philosophy, and theology, though today medical ethics is also recognized as its own discipline” (McGee & Caplan, 2007). Over time, Bioethics has grown as a discipline and its practitioners play important roles in clinical decision making, research studies, legislature, professional organizations, and community activities; in academia, government, non-governmental organizations and industry. Events that have shaped the development of Bioethics as a discipline over time are outlined in Table 1. One of the major applications of applied ethics in healthcare is to provide guidelines for assuring quality in healthcare through its assessment. Healthcare quality encompasses a variety of issues, each with ethical implications; therefore

ethical assessment issues will as well encompass several issues from a variety of angles. The goal of quality assessment in health care is continuous improvement of the quality of services provided for patients and populations and of the ways and means to produce these services (Chattopadhyay, 2009). Although several ethical principles have been defined (in addition to the main principles) such as: Trusts in relationships; Veracity; Fidelity; Avoidance of killing; Gratitude; and Reparation, that are very important in healthcare action, decision making, quality assurance and policy making. The four main principles of Bioethics are: Autonomy, Beneficence, Non-maleficence, and Justice. These principles and their applications are outlined in Figure 1.

Table 1. Developments in Bioethics – A timeline Year

Development

Resource for Further Reading

1947

The Nuremberg Code

http://ohsr.od.nih.gov/guidelines/nuremberg.html

1964

The Declaration of Helsinki

http://ohsr.od.nih.gov/guidelines/helsinki.html

1966

Animal Welfare Act

http://www.aphis.usda.gov/animal_welfare/awa.shtml

1974

National Research Act

http://www.hhs.gov/ohrp/irb/irb_introduction.htm

1979

The Belmont Report

http://ohsr.od.nih.gov/guidelines/belmont.html)

2009 (update)

US Code of Federal Regulations Title 45 Public Welfare Part 46 Protection of Human Subjects

http://www.hhs.gov/ohrp/humansubjects/guidance/45cfr46.htm

1995

National Bioethics Advisory Commission- Ethical and Policy Issues in International Research: Clinical Trials in Developing Countries

http://bioethics.georgetown.edu/nbac/human/overvol1.html

1997

International Conference on Harmonization: Guidance Documents

http://www.fda.gov/regulatoryinformation/guidances/ucm122049.htm

2002

Council for International Organizations of Medical Sciences (CIOMS) guidelines

http://www.cioms.ch/publications/layout_guide2002.pdf

2002

Nuffield Council on BioethicsThe Ethics of Research Related to Healthcare in Developing Countries

http://www.nuffieldbioethics.org/fileLibrary/pdf/errhdc_fullreport001.pdf

2003

HIV Trials Prevention NetworkEthics Guidance for Research

http://www.hptn.org/web%20documents/EWG/HPTNEthicsGuidanceFINAL15April2003.pdf)

95


CHALLENGES TO QUALITY OF HEALTHCARE Healthcare systems across the world generally do not provide consistent, high quality medical care to all people in respective nation states. Understanding of disease process, various determinants of disease and advancements in available treatments, medical technology, shifts and drifts

in disease patterns along with emerging and reemerging infections across the world have made the response to health challenges more complex and have modified the needs of public health; especially prevention of diseases which has become a major challenge. The nature of this change was captured well by the Institute Of Medicine’s (IOM) report: Crossing The Quality Chasm: A New Health System For The 21st Century in the United

Figure 1. Principles of Bioethics and their application

96


States, a report that is perhaps valid for the whole world. The Healthcare delivery system has fallen far short in its ability to translate knowledge into practice and to apply new technology safely and appropriately. There exists a large gap between the quality of the current healthcare and the quality of healthcare that can be achieved even under the current socio-economic-medical-technological environment. A number of factors have combined to create this chasm. Medical science and technology have advanced at an unprecedented rate during the past half-century. In tandem has come growing complexity of health care, which today is characterized by more to know, more to do, more to manage, more to watch, and more people involved than ever before. Faced with such rapid changes, the nation’s health care delivery system has fallen far short in its ability to translate knowledge into practice and to apply new technology safely and appropriately. And if the system cannot consistently deliver today’s science and technology, it is even less prepared to respond to the extraordinary advances that surely will emerge during the coming decades. The public’s health care needs have changed as well. Americans are living longer, due at least in part to advances in medical science and technology, and with this aging population comes an increase in the incidence and prevalence of chronic conditions. Such conditions, including heart disease, diabetes, and asthma, are now the leading cause of illness, disability, and death. But today’s health system remains overly devoted to dealing with acute, episodic care needs. There is a dearth of clinical programs with the multidisciplinary infrastructure required to provide the full complement of services needed by people with common chronic conditions.

The health care delivery system also is poorly organized to meet the challenges at hand. The delivery of care often is overly complex and uncoordinated, requiring steps and patient “handoffs” that slow down care and decrease rather than improve safety. These cumbersome processes waste resources; leave unaccountable voids in coverage; lead to loss of information; and fail to build on the strengths of all health professionals involved to ensure that care is appropriate, timely, and safe. Organizational problems are particularly apparent regarding chronic conditions. The fact that more than 40 percent of people with chronic conditions have more than one such condition argues strongly for more sophisticated mechanisms to coordinate care. Yet health care organizations, hospitals, and physician groups typically operate as separate “silos,” acting without the benefit of complete information about the patient’s condition, medical history, services provided in other settings, or medications provided by other clinicians (IOM, 2001).

GLOBAL HEALTHCARE AND ETHICS Despite much discussion, biomedical research and healthcare quality, administration and assessment in many developing countries remains poorly regulated. This is attributed to the lack of a well defined and uniform set of ethical regulations/ guidelines (Chattopadhyay, 2009). The purpose of setting up ethical guidelines is to strengthen the pursuit of improvement of quality of healthcare by means of quality assessment practices and to create ethical grounds for audit. The greatest challenge in adopting uniform global ethical standards stems from inequalities in resource and wealth distribution across populations throughout the world. Additionally, the role of social determinants of health plays a larger role in determining the health of populations compared

97


to individual medical treatments. The inertia associated with social forces is also greater, and so is the potential for large scale impacts, making it necessary to develop strong ethical guidelines to harness these forces aiming to improve health of populations globally. Current guidelines on global ethical standards address ethics as applied to the individual, rather than in the larger social context. Singer & Benatar (2001) argue that unwarranted dependence on international declarations and even revisions to ethical codes like the Declaration of Helsinki will hardly result in research becoming more ethical throughout the world unless ethical capacity is truly strengthened. We believe that the same applies to improvement of healthcare quality too. Developing global ethical standards that work in impacting healthcare quality, will require defining a reasonable ‘standard of care’ (SOC) for practice and research in developing countries, redefining concepts such as fair benefits and ancillary care, encouraging community engagement and using these concepts to improve health care and foster social justice through research (Benatar & Singer, 2010). Ultimately, to overcome these challenges, the idea of international research ethics and practical global ethics will need to be broadened “to include the ethics of how institutions operate and interact, the ethics of public health and social and economic rights (based on concern for equity) and the ethics of international relations (based on solidarity) that affect whole populations” (Benatar & Singer, 2010). Furthermore, efforts to achieve the ethical goals of “in an interdependent world, where the poorest suffer most from systemic forces adversely affecting health, will need to be underpinned by promotion and achievement of solidarity with all as global citizens, and political efforts to move towards new paradigms of thinking and action that could begin to narrow widening economic

98

disparities that threaten both our humanity and global security” (Benatar & Singer, 2010).

Distribution of Scarce Health Resources In general, resources are always scarce compared to the demand for resources. Allocation of scarce resources to address healthcare quality issues can become the source of a major ethical vexation globally. Most health economists agree that some kind of public preferences should play a major role in setting criteria for distributing scarce resources. A free-market resource allocation strategy would argue in favor of adopting a distribution strategy that maximizes financial efficiency and utilization of resources by “free will” of the users. The quality-adjusted life year (QALY) is used as a preference-based measure for the outcome of health-care activities in health economic evaluative studies (Schwappach, 2002). However, a “social worth” perspective strategy would lean in favor of a distribution of resources based on social worth/ social utility of the end-user; whereas a philanthropic based perspective would endorse charity-based resource allocation. “Social value” may be derived out of the characteristics of patients and/or factors related to the characteristics of the intervention’s effect on patients’ health. Sagev (2005) proposed a set of potential guidelines that may be used for resource allocation (see below). These guidelines were based on two premises: 1. That individual well-being is the fundamental value that should guide allocation; and 2. That “interpersonal conflicts affecting well-being should be resolved in light of several conceptions of fairness, reflecting the independent value of persons and the moral significance of responsibility of individuals for the existence of interpersonal conflicts”. The general principles outlined by Sagev (2005) are:


1. The Impartiality Principle: reasons for action are agent-neutral, rather than agentrelative, apply to all agents equally; particularly, reasons for actions should be evaluated from an impartial perspective, rather from an agent-relative perspective that accords special weight to the personal aspects of agents’ life. 2. The Well-Being Principle: there is a reason to protect and enhance the well-being of persons. 3. The Equal Chance Principle: in resolving interpersonal conflicts of well-being, there is a reason to accord equal weight to the well-being of each person, by following two hierarchical sub-principles (a) there is a reason to distribute benefits or inevitable costs between all persons equally, so that each would get the maximum possible (roughly) equal benefit or bear the minimum possible (roughly) equal loss - provided that the benefit or reduction of cost for each person is significant: or (b) when this is impossible, there is a reason to give each person the highest possible equal chance to be preferred. 4. The Importance Principle: the strength of the reason provided by the Well-Being Principle depends on the importance of the interest at stake and the conjectured probabilities concerning the possible effects of the considered action or inaction on it (positively or negatively). Assuming equal probabilities, the more important is the interest, the stronger is the reason to protect it. In resolving interpersonal conflicts, there is, therefore, a reason to prefer the person who would otherwise suffer the most severe harm or the person who could be benefitted most significantly. 5. The Substantial Difference Principle: the reason provided by the Importance Principle prevails over the reason provided by the Equal chance Principle if, assuming that

all relevant probabilities are equal, there is a substantial gap in the importance of the competing interests. 6. The Principle of Fairness: Responsibility: when an interpersonal conflict requires a choice between the well-being of individuals, there is a reason to prefer a person who is not responsible for the existence of the conflict to a person who is and a person who is less responsible to a person who is more responsible.”

Distribution of Scarce Medical Interventions Allocation of very scarce medical interventions such as organs and vaccines is a persistent ethical challenge all over the world and is becoming an important issue as globalization and medical tourism increases in scope. Although several principles and strategies have been developed to allocate medical interventions, for example: treating people equally, favoring the worst-off, maximizing total benefits, and promoting and rewarding social usefulness; “no single principle is sufficient to incorporate all morally relevant considerations and therefore individual principles must be combined into multi-principle allocation systems” (Persad et al., 2009). Simple principles may include: 1. Treating people equally (random allocation or first come first serve); 2. Favoring the worst-off: prioritarianism (sickest first/ youngest first); 3. Maximizing total benefits: utilitarianism (Number of lives saved/ prognosis or life-years saved); and 4. Promoting and rewarding social usefulness (instrumental value/ reciprocity). Multi-principle systems have also been used such as; 1. UNOS points systems for organ allocation in the USA (includes Firstcome, first-served; sickest-first; prognosis); 2. QALY allocation (includes Prognosis; excludes save the most lives; 3. DALY allocation (includes Prognosis; instrumental value; excludes save the

99


most lives); and 4. Complete lives system (includes Youngest-first; prognosis; save the most lives; lottery; instrumental value, but only in public health emergency) (Persad et al., 2009).

tion; post surgical expectations about recovery; how the patients may feel during recovery; and when they should contact the clinician; what they can eat and drink; any restrictions and alarming signs they should look for.

ETHICAL HANDLING OF INFORMATION AND PATIENT SAFETY

Collecting, processing and assessing information is a critical activity in (healthcare systems -) clinical practice, research and public health activities. Information flow is two-way process where the clinician/ researcher/ public health practitioner conveys information to the patient/ research participant and/or the public, and at the same time, assesses the responses to questions and dialogues that are generated. Several issues such as consent, capacity, disclosure, voluntary action, veracity of statements, and confidentiality of information being sought and disclosed require firm guidance to comply with ethical principles (Chattopadhyay, 2009) appropriate for conduct of research, clinical management and control of healthcare quality.

Ethical handling of health information includes privacy of the information and in order to improve patient safety, it implies that patients should have participatory role in their health information. Such a role requires certain attributes that guide collection and characterization of the information i.e. the patient should have: • • • • • •

Capacity to understand the information; Consented (informed consent) to the information; Confidentiality assured about the information; Disclosure from the clinicians about use of the information; Veracity assured by the clinicians about information, procedures and tests; and Voluntariness to provide the information.

Patients should therefore be encouraged to ask questions if they have doubts or concerns; keep and bring a list of all the medicines they take; get the results (or copies) of any test or procedure conducted on them; talk to their contact clinician about which hospital or procedure would be best for their health needs; and to make sure that they understand what will happen if they need invasive tests or invasive treatment procedures such as surgery, chemotherapy etc. Even if patients fail to ask questions, all relevant fundamental and potentially important questions should be addressed by the clinician. For example, for a surgical procedure, the patients must be informed about: the exact nature of the procedure; its dura-

100

Private Health Information Protection In the U.S., The Office for Civil Rights enforces the Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule, which protects the privacy of individually identifiable health information; the HIPAA Security Rule, which sets national standards for the security of electronic protected health information; and the confidentiality provisions of the Patient Safety Rule, which protect identifiable information being used to analyze patient safety events and improve patient safety. The goal of ethical handling of information is to ensure participant protection, which is in direct compliance with the principles of ethics. In the United States, the “Privacy Rule,” is a Federal regulation under the HIPAA of 1996 that protects certain health information. Most parties subject to the Privacy Rule were required to implement the Rule’s standards and requirements by April


14, 2003. The Rule established a Federal floor of privacy protections for most individually identifiable health information by establishing conditions for its use and disclosure by certain health care providers, health plans, and health care clearinghouses. The Rule applies to all HIPAAdefined covered entities, regardless of the source of funding. Guidelines for participant protection are outlined in Figure 2.

HEALTHCARE QUALITY ASSESSMENT The process of healthcare quality improvement and assessment has involved a host of government agencies and commissions, professional organizations, insurance underwriters, corporations,

and more recently, market forces. Traditional approaches to quality control have stressed caseby-case analysis and identifying outliers, whereas newer approaches include creating practice guidelines and profiles of hospitals and physicians (Luce et al., 1994). One of the keystone concepts of health care improvement in the United States is the belief that informed choices will lead to enhanced quality. Purchasers and the public will get better care if they can choose care that meets their needs and expectations. Health plans, clinicians, and institutions will provide better care to attract more patients, especially by appealing to the public’s ability to choose and make decisions based on this information.

Figure 2. Ethical handling of Information

101


Collecting this information on quality so that it is valid and reporting it so that it is useful present many methodologic and practical challenges. In addition, there are lingering doubts about the effect that performance information has on the choices made by the public and those who purchase health insurance on their behalf. Some also question whether quality improvement efforts undertaken by plans and providers will be designed in response to this information. How, then, can quality measurement in health care transcend the evaluation of technical performance? What are those aspects of health care and health caring that matter to people, and how can they be measured (Eisenberg, 2001)? Donabedian (1988, 2003) suggested three headings under which assessment of quality of care could be conducted: structure, process, and

outcome. Although there is no clear separation between the three domains, they are linked in ways such that structure influences processes that influence outcomes. This linking may involve different probabilities with which they may impact outcomes. The three domains may include several factors such as those outlined in Figure 3. These domains are very helpful in quality assessment, for example, in diabetic care; the following attributes may be used in quality assessment.

Structure Attributes Patient access to building, parking facilities, location and staffing of information desk, adequate computer resources and access, patient seating arrangement, clinical equipment procurements and maintenance, clinician rooms and their maintenance etc.

Figure 3. Components of Donabedian matrix for healthcare quality assessment (Adapted from Donabedian, 1988, 2003)

102


Process Attributes Training for physician, non-physician clinical and non-clinical staff, Expert team visits, adequate computer software utilization, and training for its use, data recording and data management, information given to patients, confirming, scheduling and re-scheduling of appointments, notifying patients and clinicians about any changes in plans, establishing protocols for adequate communication and interaction with laboratory for following up on tests, and their results and their reporting, hygiene and cleanliness of equipment, rooms and other quality control issues etc.

Outcome Attributes Clinical Outcomes: Effective glycemic control, increased rates of retinal screening, increased rates of foot-care examination and education, increased rates for testing for microalbuminurea, increased testing for HbA1c, and reduced HbA1c levels, Laboratory test follow ups, clinical follow ups, patient re-visit rates, return rates, patient satisfaction through interviews/surveys, physician, non-physician clinical and other staff satisfaction. Ethical guidelines to assess healthcare quality must also ensure that structure, processes and appropriate outcomes are correctly assessed when

evaluating healthcare quality. Figure 4 outlines the possible ethical guidelines check list that can be used by Ethics Boards for the quality assessment of health care

Healthcare Quality Improvement Healthcare quality improvement has been defined by Batalden & Davidoff (2007) as “the combined and unceasing efforts of everyone—healthcare professionals, patients and their families, researchers, payers, planners and educators— to make the changes that will lead to better patient outcomes (health), better system performance (care) and better professional development (learning)”. The IOM (2001) suggested six goals that healthcare systems should adopt to guide their gains to “far better at meeting patient needs”, emphasizing that “the health care system must be responsive at all times, and access to care should be provided over the Internet, by telephone, and by other means in addition to in person visits”. These six aims for improvement of healthcare quality are that the healthcare systems should be: • •

Safe (i.e. avoid injuries to patients from the care that is intended to help them); Effective: (i.e. provide services based on scientific knowledge to all who could ben-

Figure 4. Ethical guidelines for the quality assessment of health care

103


•

•

•

•

efit, and refrain from providing services to those not likely to benefit); Patient-centered (i.e. provide care that is respectful of and responsive to individual patient preferences, needs, and values, and ensure that patient values guide all clinical decisions); Timely (i.e. reduce waits and sometimes harmful delays for both those who receive and those who give care); Efficient (i.e. avoid waste, including waste of equipment, supplies, ideas, and energy); and Equitable: (i.e. provide care that does not vary in quality because of personal characteristics such as gender, ethnicity, geographic location, and socioeconomic status).

Although it is not possible to specify a complete set of rules that would be necessarily useful of applicable universally, the IOM (2001) formulated a set of ten simple rules (or general principles) to inform efforts to redesign the health system to help improve healthcare quality. These ten “rules” state that: 1. Care should be based on continuous healing relationships. 2. Care should be customized according to patient needs and values. 3. The patient should be the source of control. 4. Knowledge should be shared and information should flow freely. 5. Decision making should be evidence-based. 6. Safety should be a system property. 7. Transparency should be necessary. 8. Needs should be anticipated. 9. Waste should be continuously decreased. 10. Cooperation among clinicians should be a priority.

104

Overall, reducing risk and ensuring safety require greater attention to systems that help prevent and mitigate errors. Whereas government funding support would be pivotal to improve healthcare quality, it is critical that leadership from the private sector, both professional and other health care leaders and consumer representatives, be involved in all aspects of this effort. Incorporation of advances in information technology and establishment of fair payment systems for compensating clinicians for good patient management should be the cornerstone for such efforts.

SUPERVISION OF ETHICAL GUIDELINES FOR HEALTHCARE QUALITY ASSESSMENT Healthcare Quality Supervision Healthcare quality improvement, at any level, may it be implemented (overall systems level, at a hospital level, or small practice level), should generally be supervised by a knowledgeable team. The supervision team should be cognizant of the knowledge systems involved in improvement, some characteristics of which was suggested by Batalden & Davidoff (2007). The knowledge systems should include: 1. Generalizable scientific evidence (i.e. controls and limits context as a variable; tests hypotheses) 2. Particular context awareness (i.e. characterizes the particular physical, social and cultural identity of local care settings e.g., their processes, habits and traditions) 3. Performance measurement (i.e. assesses the effect of changes by using study methods that preserve time as a variable, use balanced measures [range of perspectives, dimensions], analyze for patterns)


4. Plans for change (i.e. describes the variety of methods available for connecting evidence to particular contexts) 5. Execution of planned changes (i.e. provides insight into the strategic, operational and human resource realities of particular settings (drivers) that will make changes happen). Supervision of ethical aspects of healthcare quality usually lies with the Ethics Board or institutional Review board of the local institutions. However, as more international tie-ups of hospital chains occur, and more international research gets conducted, the role of responses of different Ethics Boards at different sites has demonstrated major differences between how these Boards conduct their functions. Gold and Dewa (2005) examined the issue of whether the current system for ethics review of multisite health services research protocols is adequate, or whether there exist alternative methods that should be considered. They reported that: (1) Investigators at different sites in a multisite project often have very different experiences with respect to the requirements and requests of the review board. Other reported problems included the waste of time and resources spent on document preparation for review boards, and delays in the commencement of research activities. (2) “There are several possible reasons why there is variability in ethics review. These include the absence of standardized forms, differences in the background and experiences of board members, the influence of institutional or professional culture, and regional thinking. (3) Given the limited benefits derived from the variability in recommendations of multiple boards and the numerous problems encountered in seeking ethics approval from multiple boards suggest that some sort of reform is in order”. Solutions to these problems could include: establishment of multisite Ethics Review Boards; standardization of documents and procedures; use of technology to facilitate ethics review; and education and certification for Ethics Review Board members. However, challenges to

making these changes could include local Board power politics and increased feeling of competitiveness that could modify the interaction between the researcher and the Review Board. The increasing number of multisite, health services research studies calls for a centralized system of ethics review. The local review model is simply not conducive to multisite studies, and jeopardizes the integrity of the research process. Centralized multisite review boards, together with standardized documents and procedure, electronic access to documentation, and training for board members are all possible solutions. Changes to the current system are necessary not only to facilitate the conduct of multisite research, but also to preserve the integrity of the ethics approval process in general Gold and Dewa (2005).

Obligation for Impartial Quality Assessment Ethical guidelines for quality assessment concern all clinicians, institutions providing health care services for patients and producers of audit services. Ethical codes of clinicians include a commitment to an obligation for continuous improvement in their professional abilities and to evaluate the methods used. For example, provision 5 of the Code of Ethics for Nurses With Interpretive Statements (American Nursing Association, 2010) states that “the nurse owes the same duties to self as to others, including the responsibility to preserve integrity and safety, to maintain competence, and to continue personal and professional growth”; and Provision 7 states that “the nurse participates in the advancement of the profession through contributions to practice, education, administration, and knowledge development.” A clinician, therefore has to maintain and increase his/her knowledge and skills, and should recommend only examinations and treatments that are known to be effective and appropriate

105


based on medical knowledge and experience. Quality aspects are included also included in ethical guidelines, for example, provision 6 of the Code of Ethics for Nurses With Interpretive Statements states that “the nurse participates in establishing, maintaining, and improving health care environments and conditions of employment conducive to the provision of quality health care and consistent with the values of the profession through individual and collective action”. This assumes that patients have the right for high quality healthcare. That the clinician is obliged to provide optimal care is also ingrained in the ethics code for example, provision 4 of the of the Code of Ethics for Nurses With Interpretive Statements states that the nurse is responsible and accountable for individual nursing practice and determines the appropriate delegation of tasks consistent with the nurse’s obligation to provide optimum patient care. Because ethical conduct and following the professional ethical guidelines is required for maintaining professional status and employment credentials, employers of health care professionals have an obligation to therefore help in creating opportunities for a health care professional’s participation in appropriate further education, the completion of which needs to be followed by assessment. Alternately the employers may facilitate the professionals’ access to such programs.

THE ASSESSOR OF HEALTHCARE QUALITY The healthcare quality assessor should be trained in and conversant with auditing techniques, quality program and operations of healthcare systems, hospitals and associated agencies, and the competitive, legal and ethical issues associated with healthcare quality assessments. Competence of the Assessor must be well-established. A guide to competence of Assessors is outlined in Figure 5.

106

Schemas for competence assessment of Assessors may include evaluation of the following. • • • • • • • • •

How are competences Assessors appointed? What competence assessment methods are followed? The method, accuracy and security of record keeping Outcomes of the competence assessment that are carried out Consistency of controller competence assessment Refresher and unusual/emergency situations training Who are subject to competence assessment? Have these subjects been assessed? Development, testing and adoption of performance objectives

Selection and Initial Qualification Criteria for selection of assessors include: education, demonstrated working knowledge, working experience, training, and assessment experience, communication/interpersonal skills, auditing skills, flexibility and performance motivation.

Demonstration of Assessor Competence Assessors should be competent to demonstrate, assess and report assessments. They can be evaluated of their competence by interviews, written examination, demonstration, casual observation, formal evaluation, documentation, attestation, verification, documenting, interviews and review of previous work. Assessors should be aware of and understand the accreditation criteria, reference documents and their application, assessment principles, practices and techniques.


Figure 5. Competence of assessors (Adapted from ILAC, 2006)

RESEARCH AND PUBLICATION ETHICS As emphasis on evidence-based clinical practice increases, clinical research is being conducted increasingly in all parts of the world in institutions and by professionals who may or may not have had experience in conduct of research and communication of its findings. Among these are international clinical trials that also involve large sums of money that may be paid to participating institutions. Ethical conduct of research must be therefore made mandatory and staff should be appropriately trained. However, healthcare quality assessment must also assess such activities. Figure 6 outlines some concepts and resources that may be tapped for more information on research ethics. A key aspect is informed consent about which some misconceptions perhaps persist: is mere obtaining of informed consent sufficient to make clinical research ethical?

Many believe that informed consent makes clinical research ethical. However, informed consent is neither necessary nor sufficient for ethical clinical research. Drawing on the basic philosophies underlying major codes, declarations, and other documents relevant to research with human subjects, we propose 7 requirements that systematically elucidate a coherent framework for evaluating the ethics of clinical research studies: (1) value-enhancements of health or knowledge must be derived from the research; (2) scientific validity-the research must be methodologically rigorous; (3) fair subject selection-scientific objectives, not vulnerability or privilege, and the potential for and distribution of risks and benefits, should determine communities selected as study sites and the inclusion criteria for individual subjects; (4) favorable risk-benefit ratio-within the context of standard clinical practice and the research protocol, risks must be minimized, potential benefits enhanced, and the potential benefits 107


Figure 6. Research and publication ethics

to individuals and knowledge gained for society must outweigh the risks; (5) independent reviewunaffiliated individuals must review the research and approve, amend, or terminate it; (6) informed consent-individuals should be informed about the research and provide their voluntary consent; and (7) respect for enrolled subjects-subjects should have their privacy protected, the opportunity

108

to withdraw, and their well-being monitored. Fulfilling all 7 requirements is necessary and sufficient to make clinical research ethical. These requirements are universal, although they must be adapted to the health, economic, cultural, and technological conditions in which clinical research is conducted” (Emanuel et al., 2000).


Several issues need to be addressed regarding communication of information such as: “What types of information do people want? Are we giving it to them? What are the characteristics of information that influence peoples’ behavior? What effect does the type of decision have on the type of information needed? Do people in different circumstances use the same information in different ways” (Eisenberg, 2001)? These issues and results of inquires, studies, improvement projects and even advertisements have to be communicated by professionals and organizations to their peers, staff, patients and their families. Publications in scientific and lay media are effective and commonly used mechanisms to communicate research findings, and other information relating to healthcare services and quality to others. The Committee on Publications Ethics (COPE) is a forum for editors and publishers of peerreviewed journals to discuss issues related to the integrity of work submitted to or published in their journals. COPE supports and encourages editors to report, catalogue and instigate investigations into ethical problems in the publication process. The COPE web site may be accessed at: http:// publicationethics.org/ for gaining further insights into the role of publication ethics and how it may impact ethical communication and conduct of research and other healthcare information. Figure 6 outlines some concepts and resources that may be tapped for more information on publication ethics. Note: The views expressed in this chapter are those of the authors and do not represent the position of NIH/ NIDCR/ United States Government.

REFERENCES American Nursing Association. (2001). Code of ethics for nurses with interpretive statements. Retrieved August 5, 2010, from http:// nursingworld. org/ ethics/ code/ protected _nwcoe 629. htm

Batalden, P. B., & Davidoff, F. (2007). What is “quality improvement” and how can it transform healthcare? Quality & Safety in Health Care, 16(1), 2–3. doi:10.1136/qshc.2006.022046 Benatar, S. R., & Singer, P. A. (2010). Responsibilities in international research: A new look revisited. Journal of Medical Ethics, 36(4), 194–197. doi:10.1136/jme.2009.032672 Chattopadhyay, A. (2009). Bioethics in oral health - Clinical and research ethics in a globalized era. In Chattopadhyay, A. (Ed.), Oral health epidemiology - Principles and practice. Jones & Bartlett. Donabedian, A. (1988). The quality of care. How can it be assessed? Journal of the American Medical Association, 260(12), 1743–1748. doi:10.1001/jama.260.12.1743 Donabedian, A. (2003). An introduction to quality assurance in health care. Oxford University Press. Eisenberg, J. M. (2001). Overview to the issue Health care quality assessment. Health Services Research. Retrieved from http:// findarticles.com/ p/ articles/ mi_m4149/ is_3_36/ ai_77577986/ Emanuel, E. J., Wendler, D., & Grady, C. (2000). What makes clinical research ethical? Journal of the American Medical Association, 283, 2701– 2711. doi:10.1001/jama.283.20.2701 ILAC. (2006). ILAC Guidelines on qualifications and competence of assessors and technical experts. (ILAC-G11:07/2006). Australia: Silverwater. IOM Committee on Quality of Health Care in America. (2001). Crossing the quality chasm: A new health system for the 21st century. Washington, DC: Institute of Medicine, National Academy Press. Larson, J. S., & Muller, A. (2002). Managing the quality of health care. Journal of Health and Human Services Administration, 25(3), 261–280.

109


Luce, J. M., Bindman, A. B., & Lee, P. R. (1994). A brief history of health care quality assessment and improvement in the United States. The Western Journal of Medicine, 160(3), 263–268. McGee, G., & Caplan, A. (2007). An introduction to bioethics - Scopes of the field and a brief history of bioethics. Retrieved from http://www.bioethics. net/articles.php?viewCat=3&articleId=1 Persad, G., Wertheimer, A., & Emanuel, E. J. (2009). Principles for allocation of scarce medical interventions. Lancet, 373(9661), 423–431. doi:10.1016/S0140-6736(09)60137-9 Schwappach, D. L. (2002). Resource allocation, social values and the QALY: A review of the debate and empirical evidence. Health Expectations, 5(3), 210–222. doi:10.1046/j.1369-6513.2002.00182.x Segev, R. (2005). Well-being and fairness in the distribution of scarce health resources. The Journal of Medicine and Philosophy, 30, 231–260. doi:10.1080/03605310590960120 Singer, P. A., & Benatar, S. R. (2001). Beyond Helsinki: A vision for global health ethics. British Medical Journal, 322(7289), 747–748. doi:10.1136/bmj.322.7289.747

110

KEY TERMS AND DEFINITIONS Ethics: Branch of Philosophy that guides moral conduct life and helps living a good life. Privacy Rule: Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule, protects the privacy of individually identifiable health information. Healthcare Quality Improvement: The combined and unceasing efforts of everyone—healthcare professionals, patients and their families, researchers, payers, planners and educators— to make the changes that will lead to better patient outcomes (health), better system performance (care) and better professional development (learning). Obligation for Impartial Quality Assessment: Ethical codes of clinicians include a commitment to an obligation for continuous improvement in their professional abilities and to evaluate the methods used.

111

Chapter 7

Systems Biology Methodologies for the Understanding of Common Oncogenetic Mechanisms in Childhood Leukemic and Rhabdomyosarcoma Cells George I. Lambrou University of Athens, Greece Maria Adamaki University of Athens, Greece Eleftheria Koultouki University of Athens, Greece Maria Moschovi University of Athens, Greece

ABSTRACT Human neoplasias are considered lethal in the majority of cases. Cancer does not discriminate between age, gender, or race. Besides the detrimental health effect caused by the disease, it also has a great impact on the quality of life of those affected by it. One of the main characteristics of cancer is that it deprives the patient from their human dignity, at least compared to other diseases. For example, more people die every year from coronary disease than from cancer, but the latter is considered more severe due to the slow, painful, and deteriorating effect it has on the human body. However, in the case of childhood hematologic neoplasias, progress in the field of treatment has taken immense steps forward, and five-year survival reaches 80%. At the same time, several new therapeutic agents, such as glucose analogs or telomerase inhibitors, are currently being tested in clinical trials, and several others, such as kinase or proteasome inhibitors, are already in the market. Yet the most effective therapies are the ones performed with classical chemotherapeutics, which are non-specific and more aggressive. DOI: 10.4018/978-1-61350-120-7.ch007


Systems Biology Methodologies for the Understanding of Common Oncogenetic Mechanisms

One of the main concerns of modern life is the level and quality of health services, and at the same time, there is a great debate on the nature of health services provision, i.e. public or private. One of the main burdens of health services is drug administration. Especially in the case of cancer treatment, drug administration is extremely expensive, both for the patient and the health provider. If we think in terms of the Western world, then this is really a matter of finance, and the majority of the population can afford, either by private or state health insurance, the cost of a cancer treatment. However, if we see things in terms of developing countries or under-developed countries, where finances and health insurance are limited, then cancer treatment can not even be considered as an option for the majority of the population. The etiology for the high cost of cancer treatments comes from the process of drug development, as the latter takes almost 10 years of research and about 800 million dollars for a drug to reach the market. Most of the efforts in drug discovery are based mainly on the method of trial and error. This, in turn, is mainly due to the lack knowledge on the mechanisms underlying the disease. Therefore, let us imagine a best case scenario where oncogenesis is understood, at least in part, and drug design, effectiveness, and side effects could be resolved by simply modeling the system. This would dramatically reduce drug costs, with positive consequences not only for treatment, but also on the social level. Drugs would become much more affordable, hence curing and improving the lives of more people. This hypothetical, best-case scenario should point scientists to move towards the direction of attempting to make scientific endeavors for the social benefit. Systems biology is a discipline that does indeed move towards that direction. Of course, as most things in life, the use of systems biology could be used in a dual manner, that is, for social benefit or profit.

INTRODUCTION The Definition and Necessity of Systems Biology in Biological Phenomena Systems biology can be considered a relatively young discipline. It was in the advent of the 21st century that the first coordinated attempts became reality. However, the term had been previously mentioned both as a necessity and as a discipline. We couldn’t phrase it in better words than Mihajlo Mesarović in 1968: …in spite of the considerable interest and efforts, the application of systems theory in biology has not quite lived up to expectations…one of the main reasons for the existing lag is that systems theory has not been directly concerned with some of the problems of vital importance in biology… The real advance in the application of systems theory to biology will come about only when the 112

biologists start asking questions which are based on the system-theoretic concepts rather than using these concepts to represent in still another way the phenomena which are already explained in terms of biophysical or biochemical principles. …then we will not have the application of engineering principles to biological problems but rather a field of systems biology with its own identity and in its own right… (Mesarovic, 1968) (adopted from the book Systems Biology-Dynamic Pathway Modeling by Olaf Wolkenhauer (Wolkenhauer, 2010)). But in order to further define the field of systems biology we should refer to an older reference from Henri Poincaré: …life is a relationship among molecules and not a property of any molecule…Science is built up of facts, as a house is with stones. But a collection of facts is no more a science than a heap of stones is a house… (Wolkenhauer, 2010).


We think that there is no better way for describing the necessity for a systems approach to biological phenomena. Natural sciences in general have matured with time especially with the integration of mathematics. For example, the disciplines of physics and chemistry have benefited from the mathematical discipline and matured enough to create a common language for describing phenomena. Similarly, biological sciences could greatly benefit by integrating knowledge from the mathematical and engineering disciplines. Biology deals with some very complicated phenomena such as embryogenesis, neurophysiology, oncogenesis and others. Oncogenesis is of crucial importance not only from the biological point of view but also from the disease perspective. Tumors are considered to be lethal in the majority of cases and can deprive the affected individual from their quality of life. Despite the research efforts in the field most of the mechanisms underlying this disease are largely unknown. Besides the biological effects that are under investigation in active research, there are several very important aspects of social life that are also affected. Cancer does not discriminate between age, gender, or race. Besides the detrimental health effect caused by the disease, it also has a great impact on the quality of life of those affected by it. For example, more people die every year from coronary disease than from cancer but the latter is considered more severe due to the slow, painful and deteriorating effect it has on the human body. Perhaps the most crucial concern of modern life is the extent and quality of health services while at the same time there is a great debate on the nature of health services provision i.e. public or private. One of the great burdens of health services is drug administration. Up-to-date, chemotherapeutics consist of the main arsenal against the disease. Another important aspect is treatment. From the economics point of view cancer treatments are extremely expensive both for the patient and the health provider. If we think in terms of the western world, then this is really a matter of finance and

the majority of the population can afford, either by private or state health insurance, the cost of a cancer treatment. However, if we see things in terms of developing countries or under-developed countries, where finances and health insurance are limited, then cancer treatment can not even be considered as an option in the majority of cases. The etiology for the high cost of cancer treatments comes from the process of drug development, as the latter takes almost 10 years of research and about 800 million dollars for a drug to reach the market. Most of the efforts in drug discovery are mainly based on the method of trial and error. This is due to the lack knowledge on the mechanisms underlying the disease. Yet, classical chemotherapeutics are nowadays affordable since for most of them patent rights have expired and generic forms can be produced. However, the problem with classical therapy is the side effects. Unfortunately, effectiveness is accompanied with non-specificity. Classical chemotherapeutics do not discriminate between healthy and diseased cells. They rely on the fact that cancer cells are more active both metabolically and transcriptionally, especially on the level of cell signaling. Therefore, let us imagine a best case scenario where oncogenesis is understood, at least in part, and drug design, effectiveness and side effects could be resolved by simply modeling the system. This would dramatically reduce drug costs with positive consequences not only for treatment but also on the social level. New drugs would become much more affordable, hence curing and improving the lives of more people. Therefore, a question arises when it comes to systems biology and cancer. How can the discipline of systems biology help improve treatments and help maintain the patient’s quality of life? There is no straightforward answer to this. Due to the complicated nature of biological systems, it has been impossible until the advent of personal computers to even imagine that one would be able to study biological phenomena on the systems level. Nonetheless, the term systems

113


biology is not new. As we mentioned, it appeared in the mid 60s’ and since then a considerable amount of effort has been made to model biological systems. Nowadays, it has become more evident that maturity in biology would come through its interaction with the disciplines of mathematics and engineering. In other words, a better understanding of biology can be achieved through the progress of computational and systems biology. Systems biology is still in its infancy, since orchestrated and coordinated efforts find their start at the beginning of the 21st century. Until recently, the principle of biology was in a way restricted on the level of studying several factors simultaneously. With technological progress, through the application of high-throughput methodologies such as microarrays and deep sequencing, mass production of data became possible. This gave us the ability to examine multiple factors simultaneously, bringing closer a more systems-oriented analysis of biological phenomena. As mentioned earlier, in order to improve a certain condition, one needs to first understand the phenomenon underlying it or, in other words, understand the governing rules of the system. Systems biology does exactly that, i.e., it attempts to understand and define the rules governing biological phenomena. In the case of chemotherapeutics, two routes could be followed. The first would be to understand the mechanisms that lead to a disease, i.e. the pathogenetic mechanisms, and intervene in such a way that it would correct “mistakes” in cell function or, even better, eliminate diseased cells in the case of cancer. The second route would be to develop new or improve the existing drugs specific for a disease. In the case of cancer treatment there could be an improvement of classical therapies, since these are already available and would therefore be more feasible to do so financially, as opposed to exclusively concentrating on the discovery of new drugs for treatment, which is undoubtedly an extremely expensive and timeconsuming process. In the present chapter we will analyze the workflows that can be followed in order

114

to understand the causes of oncogenesis. We will provide useful information on the computational work flow that could help investigate the process of tumor progression on a small scale, and we will present this approach through an example study of two distinct, but devastating nonetheless, childhood neoplasias. Such an approach could assist in the understanding of the mechanisms underlying malignant transformation and as a consequence have direct effects on human health practitioning and quality of life. More specifically, we will present a computational and systems biology approach to the study of common mechanisms between two childhood neoplasias, acute lymphoblastic leukemia and rhabdomyosarcoma. Childhood leukemia is the most common form of childhood malignancy, whilst on the other hand rhabdomyosarcoma is a rare type of malignancy, belonging to the primitive neuroectodermal family of tumors. We know that tumors possess distinct characteristics not only between different types of malignancy but also between two patients affected by the same type of tumor. Despite all this, there is enough evidence to suggest that tumors might share some common characteristics which, once discovered, would lead to: a) more efficient treatments, b) a better understanding of the origin of oncogenesis and c) the discovery of universal tumor biomarkers.

Theories of Carcinogenesis and Elucidation of Their Mechanisms In the case of cancer tumorigenesis it is the interconnection of thousands of factors acting simultaneously that bring about the malignant phenotype. Several theories have been suggested for the occurrence of tumors, the most prominent one being the accumulation of mutations throughout an individual’s lifetime (Knudson, 2001); yet other equally significant theories have been suggested, such as the multistage oncogenesis theory (Moolgavkar, 1978; Ritter, Wilson, Pompei, & Burmistrov, 2003). However, in the case of child-


hood neoplasias these models/theories cannot be applied directly since the elapsed time of exposure to mutagenesis is not adequate to explain the appearance of cancer. There are several hypotheses regarding childhood neoplasias, as for example in the case of childhood leukemia it has been suggested that cytogenetic aberrations might play a role in the transformation of malignancy. Another interesting theory is that tumors are directly connected to the industrialization of modern life and it has been suggested that cancer is not actually a disease but an evolutionary adaptation of the cells in order to survive certain environmental changes/ threats. The origins and etiology of cancer are still being investigated. It is therefore evident that in order to be able to improve a certain condition, one first needs to understand the phenomenon or, in other words, to understand the governing rules of the system. Systems biology works towards understanding and defining the rules governing biological phenomena. Here we suggest two routes for the accomplishment of this purpose. The first route would be to understand the mechanisms that lead to disease, i.e. the pathogenetic mechanisms, and intervene in such a way that would try to correct “mistakes” in cell function or, even better, eliminate diseased cells in the case of cancer. This route could be further divided into two subcategories: a) the discovery of differences between tumor subtypes or differences in the reactions of tumor cells to specific treatments, one of the principal approaches used so far, and b) the discovery of common traits between different types of tumors, not only between subtypes of the same family of tumors but also between different types, both histologicaly and phenotypicaly, an approach not often referenced in literature. The second route would be to develop new drugs specific for a disease, or improve the existing ones. In the case of cancer treatment there could be an improvement of classical therapies, since these are already available, and would therefore be more feasible to do so financially, as opposed to exclusively

concentrating on the discovery of new drugs for treatment, which is undoubtedly an extremely expensive and time-consuming process. We think that the common aspects that characterize tumors irrespectively of type or subtype are of major scientific interest. It is possible that the mechanics leading a cell to become malignant could be similar among certain cancer types. For example, a common characteristic of almost all types of tumors is the known Warburg effect described by Otto Warburg in 1924. This simple observation remains until today a fundamental trait of tumor biology i.e. that tumors perform a shift from oxidative phosphorylation to aerobic glycolysis for their energy needs (Warburg, Posener, & Negelein, 1924). This is a characteristic that all tumors have in common and pharmaceuticals exploiting this trait are already in clinical trials. We believe that it is possible for a common oncogenetic mechanism to exist and that its elucidation could provide a wide range of weaponry in the fight against cancer. Not only would it be possible to gain a more indepth understanding of tumor biology, it would also open up new ways for treating cancer.

Childhood Neoplasias and Common Mechanisms Until recently, scientific approaches concentrated on examining several factors simultaneously, which only revealed part of the picture. Oncogenesis is a highly complicated process and one of the tools that current research has in its arsenal is computational and systems biology. In this respect, we are now able to investigate possible common mechanisms that govern different neoplasias. In the present chapter we will analyze and present workflows that can be followed in order to understand the causes of oncogenesis. We will do so using two childhood neoplasias, both of different cellular origin and of different phenotype, as an example. These are acute lymphoblastic leukemia and rhabdomyosarcoma. The first is the most common cancer of childhood cancer and the second

115


one of the rarest tumors found in children. ALL is considered to be one of the great successes of the century with cure rates that reach 80% of the cases, whereas survival rates in other neoplasias still remain very low. We use reverse engineering methodologies as an approach of finding common mechanisms of cancer progression.

ACUTE LYMPHOBLASTIC LEUKEMIA AND RHABDOMYOSARCOMA AS CASES Acute lymphoblastic leukemia (ALL) is the most frequent occurring malignancy among childhood cancers. It originates in the undifferentiated lymphoblast, which abnormally ceases develop into the mature lymphoid cell giving rise to a tumor. Acute leukemia mainly appears during childhood but it also occurs in adolescence manifesting a poor prognosis regardless of age. Progress in childhood leukemia has been immense the last decades with an overall survival rate exceeding 75% (Carroll et al., 2003). The prognostic factors have also been well characterized for childhood leukemias; these include white blood cell count at presentation (diagnosis), age and gender, as well as certain chromosomal aberrations (Carroll, et al., 2003). Treatment of childhood leukemias is successful for the majority of patients, mainly due to the use of classical chemotherapeutics. Recently individualized treatments have also been applied, as in the case of BCR/ABL positive (also known as Philadelphia positive (Ph+)) leukemias, using imatinib mesylate, a new agent specific for the particular gene fusion.. This represents a great example of how childhood leukemia can benefit from individualized therapy. On the other hand, rhabdomyosarcoma (RMS) is a rare childhood cancer representing 5-8% of all tumors emerging during childhood. However, the sarcomas of head and neck comprise 12% of all childhood neoplasias (Chadha & Forte, 2009) with an incidence rate of 250 cases every year in

116

the United States. Rhabdomyosarcoma is a malignant tumor of mesenchymal origin and it is within the category of small blue cell tumors, including neuroblastoma and primitive neuroectodermal tumors (Chadha & Forte, 2009). The embryonal form is most common at birth whereas the alveolar form peaks in childhood and adolescence (Toro et al., 2006). Embryonal rhabdomyosarcoma, characterized by spindle cells and botryoid form, is associated with better prognosis (Leaphart & Rodeberg, 2007). As in the case of childhood leukemia, many alveolar rhabdomyosarcomas present chromosomal translocations, such as the PAX3-FKHR, which is an indicator of bad prognosis, since it is highly associated with metastasis (Chadha & Forte, 2009; Davicioni et al., 2009; Oda & Tsuneyoshi, 2008; Sorensen et al., 2002). Rhabdomyosarcoma is rarely cured by one method of treatment alone. Therapy protocols include chemotherapy and either surgery or radiotherapy. In general the 5-year survival rates are in the extent of 97% in embryonal non-orbital, non-parameningeal head and neck cases, and 67% in the alveolar cases (Pappo et al., 2003). The cell of origin is considered to be the myoblast or cells that will form the skeletal muscle. These cells are different from the smooth cells that line the intestinal tract. Theoretically, rhabdomyosarcomas can emerge at any part of the body that has skeletal muscle but it originates mainly in the head and neck.

COMMON CHARACTERISTICS BETWEEN ALL AND RMS CELLS A common aspect between ALL and RMS is that both comprise of cells that are undifferentiated, immortal and with the potential to divide infinitely. This means that proliferation takes place in an uncontrollable pattern. Interestingly, it has been reported that rhabdomyosarcoma can be present in the bone marrow of patients presenting a leukemic image, without the presence of a primary tumor (Morandi, Manna, Sabattini, & Porcellini, 1996;


Parham, Pinto, Tallini, & Novak, 1998; Putti et al., 1991; Sandberg, Stone, Czarnecki, & Cohen, 2001). In addition, it has been reported that there is a possibility for presentation of secondary malignancy after successful treatment of a primary (therapy-related leukemia). For example, following the successful treatment of a paravertebral embryonal rhabdomyosarcoma (ERS), a patient developed acute T-lymphoblastic leukemia (Kaplinsky et al., 1992). The question that could be posed here is whether this development of the secondary tumor originated from cells of the primary tumor, as in the case of therapy-related leukemia, or whether there were “dormant” leukemic cells in the bone marrow which got triggered to proliferate uncontrollably after rhabdomyosarcoma chemotherapy. In the case where cells from one type of cancer develop into another, it probably means that the tumor cell possesses two traits: first it is able to migrate and second it is able to differentiate to another type of cell manifesting stem cell properties. This is in accordance with the stem cell theory of cancer origin, as cancer stem cells keep their ability to differentiate, migrate and even give rise to a new malignancy with almost totally new traits. Furthermore, it has been reported that therapy-related leukemia can occur due to the use of chemotherapeutics (Dedrick & Morrison, 1992; Park & Koeffler, 1996). This phenomenon has not been thoroughly investigated and the mechanisms behind it are still obscure. However, it points out the fact that carcinogenesis is a complicated process, implicating a number of mechanisms and not just single events in a cell’s life time. If we were to accept the hypothesis that two different tumor cells can co-exist in different locations as true, then we could accept the notion that stem cells play the main role in carcinogenesis and tumor growth. On the other hand, an interesting report by Kelly et al. showed that, at least in part, the presence of cancer stem cells is not necessary for tumor growth to be sustained (Kelly, Dakic, Adams, Nutt, & Strasser, 2007).

By looking into the developmental progress of the aforementioned cell types, both seem to originate from the embryonic mesoderm. Myoblasts originate from the dorsal (paraxial) mesoderm whereas blood cells derive from the lateral mesoderm, which gives rise to the splachnic mesoderm and this, in turn, to the hemangioblastic tissue. Further on, paraxial mesoderm gives rise to the un-segmented presomitic mesoderm, formed during gastrulation, and mesoderm generated from the primitive streak. One of the earliest genes expressed in the paraxial mesoderm is T (branchyury). This is also an embryonic transcription factor that is expressed in a gradient across developmental sites derived from the primitive streak and continuously expressed in the notochord (Sewell & Kusumi, 2007). The FGF (Fibroblast Growth Factor) and RA (Retinoic Acid) pathways are the main routes followed in early somatic development (Sewell & Kusumi, 2007). Another important signaling pathway that has been found to participate in developmental as well as oncogenic pathways is the NOTCH pathway. Notch genes encode transmembrane receptors. The human genome contains four NOTCH receptor homologues. In somite stage, NOTCH1 is expressed across the presomitic mesoderm, where mutations have been found to cause defects in somitic segmentation and anterior-posterior polarity (Dunwoodie et al., 2002; Kusumi et al., 1998). On the other hand, during embryogenesis, blood cells originate from two sites: the first is thought to be the ventral mesoderm near the yolk sac, which gives rise to the intra-embryonic hematopoietic precursors, whereas the hematopoietic cells that last throughout the entire life time of an organism are derived from the mesodermal area surrounding the aorta (Godin & Cumano, 2005). From the point of mesoderm differentiation, strictly orchestrated regulation of various genes leads to two similar cell types with different functions and roles in the body. This is the result of a complicated network of gene regulation and expression. It is easy to assume that aberrations in the regulatory

117


networks underlying development would lead to tumor cells. Developmentally, there are several factors that affect gene regulation in order for differentiation to take place. Tumors can be as different as the patients that carry the disease. It is these differences that make tumors so hard to cure, since therapies have different results on different patients. To our present knowledge there are not many reports trying to identify common gene expression profiles between the two aforementioned tumors. An interesting reference highlights the fact that stem cells are probably found in different types of tumors, thus suggesting that stem cells are implicated in the etiology of tumor maintenance and growth (Nicolis, 2007). However, there is evidence that even normal, already differentiated cells, can be transformed to tumorigenic ones (Nicolis, 2007). Based on this observation, there was a case reporting a child manifesting 5 different tumor types simultaneously (Perilongo et al., 1993). It could be that stem cells originating from the same organism possess similar defects but enough alterations to be able to give rise to five different types of tumor. The discovery of similar or opposing gene expression profiles may lead to the understanding of a common tumor origin, if such exists. Another point on which attention should be drawn is the regulation of genes through transcription factors (TFs). Knowledge of gene regulatory networks is considered to be of crucial importance for the understanding of diseases such as cancer, as it may lead to new therapeutic approaches (Chang, Nagarajan, Magee, Milbrandt, & Stormo, 2006). The knowledge of common transcriptional regulatory networks could potentially lead to a universal treatment for diverse diseases such as cancer, and it is through this possibility that the need for computational methods in the study of carcinogenesis becomes apparent.

118

IN VITRO CELL MODELS FOR THE STUDY OF COMMON MECHANISMS The study of the mechanisms underlying carcinogenesis or tumor progression or resistance to chemotherapeutics requires biological systems that would resemble these processes in vivo. There are several models available both in experimental praxis as well as those reported in the literature. One of the main approaches utilized is the use of immortalized cell lines, taken as biopsies either from human or other species, and transformed to divide indefinitely in laboratory conditions. Cell lines manifest several advantages and disadvantages. Among the advantages, we value the possibility for repeatability in experimental procedures, the ease of use in daily laboratory praxis and the feasibility of use from an economical perspective. On the other hand, cell lines do not accurately represent the cellular systems, the pathogenesis or the phenotype as it is presented in vivo. However, as far as the ethical part of experimentation goes, cell lines outreach all other study models. Ethical considerations have been raised due to the uses of human specimens or laboratory animals. There is also a general tendency for an increase in the use of cell lines as models of study and the decrease of other types of biological systems, although patient samples and animal models are irreplaceable in certain cases. Here we use two cell lines as our reference systems. We examine the experimental and analysis steps taken in order to elucidate common patterns of cellular function between these two systems. The two cell lines are the T-cell acute lymphoblastic leukemia CCRF-CEM cell line and the rhabdomyosarcoma TE-671 cell line. Both were obtained from the European Collection of Cell Cultures (ECACC). The CCRF-CEM cell line, a CD4+ (Miranda, Wolf, Pichuantes, Duke, & Franzusoff, 1996) and CD34+ presenting cell line (Naujokat et al., 2000), was initially obtained from the peripheral blood of a 2-year-old Caucasian female diagnosed with lymphosarcoma, which later on progressed


to acute leukemia (Foley et al., 1965). The child underwent irradiation therapy and chemotherapy prior to obtaining the cell line. Although remission was achieved at various stages, the disease progressed rapidly (Foley, et al., 1965). The cell line has been observed to undergo minor changes after long-term culture, except for the presence of dense granules in the nucleoli (Uzman, Foley, Farber, & Lazarus, 1966) and to manifest autocrine catalase activity, an important regulator of the cell line’s growth(Sandstrom & Buttke, 1993). The fact that these cells have been therapeutically treated before collection makes them an ideal model for the investigation of drug resistance; as a matter of fact, it has been previously reported that the particular cell line manifests resistance to several chemotherapeutic agents, e.g. glucocorticoids (Lambrou et al., 2009). The TE-671 was initially reported to be obtained from a cerebellar medulloblastoma, before irradiation therapy, of a six-year old Caucasian female (McAllister et al., 1977) and was characterized later on (Friedman et al., 1983). Nowadays we know that this cell line is parental, if not identical, to the RD (McAllister, Melnyk, Finkelstein, Adams, & Gardner, 1969) rhabdomyosarcoma cell line. However, several reports still refer to this cell line as medulloblastoma (Chu, Wong, & Yow, 2008; Yeung, An, Cheng, Chow, & Leung, 2005). Based on our positions stated above, these cells share several common characteristics despite their obvious differences. Both cell types are primarily derived from the mesoderm, where cells stopped differentiating, probably at an early stage, before maturing into their final phenotype. Physiologically, these cells would have matured and differentiated into the cells constituting the blood and muscle cells. Also, at some unknown point, normal differentiation ceased and the cells started accumulating uncontrollably, i.e. they became malignant. From that point on, until the first manifestation of symptoms of malignancy, there is a lack of knowledge of the mechanistical steps leading to oncogenesis. In other words, we

cannot acquire ant form of knowledge for the stages that involve the emergence of the blasts and their proliferation dynamics up to the point that they are clinically presented. This represents another good reason for studying such systems using in vitro models, since in vitro models allow the creation of proliferation and growth models in long term cultures, something that is impossible to achieve with ex vivo cells or animals. Our point here is whether two distinct cell types, destined to perform different functions, manifest similar mechanisms of growth and progression due to their malignant character. In other words, is the phenotypic character of the cells originally inherent through their developmental histoire, or is it due to the malignant character, acquired at some point of development, that determines cell proliferation, growth, gene expression and regulation. For the present, we will focus on the differential expression profile underlying the two cell lines and on the computational approach of the aspect surrounding a common expression profile. In particular, we will look at groups of genes based on their function and on their regulation. For example, it is expected for two different cell lines to manifest differential gene expression. Yet, what are the genes that are commonly expressed? We mainly focus on the similarities between the two cell types. Based on these observations we could narrow down our search for common mechanisms and focus on groups of genes such as the active developmental genes, the molecules secreted from both cell types, genes that sustain growth and genes involved in the cell cycle. We will present the use of high throughput and computational approaches and try to elucidate which genes (within a certain group of genes) are typical for each cell type, but at the same time genes specific for one cell type appear in the other. We expect to find some genes active in both cell types. We will attempt to find similarities in tumor progression, cell cycle and secretory extracellular signaling. We will try to elucidate the possible common regulatory mechanisms for the two

119


cell types based on their expression profile and, investigate whether there are pathways common to both cell types; the latter could be modeled as a means of investigating how gene function is implicated in the networking observed in all biological systems, especially in cancer. Finally, we will attempt to reach some conclusions regarding the origins of tumorigenesis. In other words, from the computational analysis performed we will try to deduce whether stem cells play a role in tumor progression. Several studies have been occupied with the detection of cancer germ line genes (CGGs) both in pediatric sarcomas (Jacobs et al., 2007) and pediatric brain tumors (Jacobs et al., 2008). The discovery of global antigens for tumor vaccines could provide salvation for childhood malignancies (Jacobs, et al., 2007). Some studies have reported the appearance of common antigens in Ewing sarcoma/primitive neuroectodermal tumor (EWS/PNET) and lymphoblastic lymphoma (Mhawech-Fauceglia et al., 2007). Rhabdomyosarcoma is included in the primitive neuroectodermal tumor types, where embryonic genes are involved in cancer progression (Blandford et al., 2006). Leukemia, and more specifically ALL, on the other hand, originates from the lymphoblast, also known to employ embryonal mechanisms for its differentiation and progression. This type of approach opens up the road to a more holistic understanding of the tumor/oncogenetic dynamics, since by finding common mechanisms of action it would be possible to better understand the origins of cancer. Different neoplasias with different characteristics are expected to possess different traits and phenotypes. However, an approach that identifies commonalities between such diverse biological systems could lead to the understanding of the common mechanisms that transform physiological cells to malignant. This could have a very positive outcome in drug development and contribute to new, more global, therapeutic approaches.

120

EXPERIMENTAL PROCEDURES AND COMPUTATIONAL WORKFLOW The analysis can be divided in two parts: the experimental part and the computational part. The experimental part determines the outcome of the computational analysis to a greater extent. This means that experimental design should be performed in such a way that it should outline, when possible, the main questions that need to be answered. This is not always the case though. There are several experimental set-ups which are mostly discovery-driven and less hypothesis-driven. In the present case we use a discovery driven approach. Below is an outline of the total workflow that can be used for such analyses: • • • • •

• •

• • • •

Cell culture and growth Allow cell proliferation to reach the desired extent Cell proliferation measurements Cell collection, at desired time points, for further processing Cell collection for: ◦âŠ¦ RNA extractions ◦âŠ¦ DNA extractions (optional) ◦âŠ¦ Protein extractions (optional) ◦âŠ¦ Cell culture supernatants for measurements of extracellular molecules and factors ◦âŠ¦ Cells for flow cytometric analyses Extractions of the aforementioned molecules Flow cytometric analyses for cell death, cell size, granulation and cell cycle distribution Computational analysis of flow cytometric data (optional) Microarray experimental design Microarray experimentation between selected samples Computational analysis of microarray data ◦âŠ¦ Signal filtering ◦âŠ¦ Background correction


◦âŠ¦ ◦âŠ¦ ◦âŠ¦ ◦âŠ¦

◦âŠ¦ ◦âŠ¦ ◦âŠ¦ ◦âŠ¦ ◦âŠ¦ ◦âŠ¦ ◦âŠ¦

Signal intensities normalization Statistical tests for differential gene expression detection Hierarchical clustering Other clustering methods (e.g. k-means, Principal Component Analysis) Transcription Factor Binding Motifs predictions (TFBMs) Chromosome distribution analyses Gene Ontology (GO) analyses Pathway participation and mapping analyses Pathway simulations (optional) Protein-protein interactions prediction (optional) Final conclusions

At this point we are able to describe the workflow in a step-by-step manner and to further explain the necessities for each of the steps taken.

Cell Culture and Growth Both cell lines were seeded at time -24h and left to grow overnight. This is a necessary step for the cells to re-adjust to their environment and avoid the stress effects caused during feeding and seeding in future measurements. After 24h (time: 0h) a sample was taken from both cell lines in order to perform measurements and every 24h thereafter. Cell lines were grown in culture media, as for example the T-Lymphoblastic Leukemia CCRF-CEM cells were grown in RPMI-1640 medium supplemented with 2mM L-Glutamine and Streptomycin/Penicillin 100 U/ml, 20% FBS at 37 oC, 5% CO2 and ~100% humidity. TE-671 cells were grown in DMEM medium supplemented with 2mM L-Glutamine, 10% FBS and 100 U/ml Streptomycin/Penicillin. Cells were allowed to grow to a population of ~1.500X103 cells/ul for CCRF-CEM and at ~80% confluence for TE-671. Cells were harvested at confluence using 0,1% trypsin (only for TE-671)

and centrifugation at 1000 rpm for 10 min. Supernatant was discarded and cells were washed with pre-warmed 1X PBS, re-centrifuged at 1000 rpm for 10 min and the cell pellet was kept for further processing.

Cell Proliferation Cell population is an important characteristic of cell type. It is important to mention that the two cell systems used are very different in the way they grow. The CEM cell line grows in suspension i.e. in a 3D-environment very closely resembling the in vivo condition of the bone marrow. On the other hand the rhabdomyosarcoma cells grow in an adherent mode, i.e. they attach to the collagenous substrate of the flasks they grow in. This immediately manifests two basic differences. In the case of leukemic cells confluence can not be directly determined since it is a question of nutrition abundance and a question of availability of space. On the other hand, RMS cells, regardless of nutrition abundance, are highly depended on space availability, since the mechanism of contact inhibition can lead them to cell death once confluence is reached. Therefore, cell counts must be regularly performed. In the present case cells were counted at the -24 h time point as well as at 0 h, 4 h, 24 h, 48 h, and 72 h after having been let to grow under normal conditions. For this purpose, 200 μl of cell suspensions were obtained from each flask and counted directly with a hematologic analyzer.

Flow Cytometry Flow cytometry is a powerful method that allows the simultaneous determination of multiple factors such as cell size, granulation, membrane markers, cell death and viability, DNA content or cell cycle distribution. In the present case we present as an example the determination of three factors: 1) cell size and granulation, 2) apoptosis and 3) cell cycle distribution. Cell cycle distribution and DNA content was determined with standard propidium

121


iodide (PI) staining, a stain with high affinity for nucleotides. When PI gets excited by an energy source it gives off fluorescence proportional to the DNA content of the cells, thus indicating the G1, G2 and S phases. Briefly, 1ml of cell suspension from each flask was centrifuged at 1000 rpm for 10 min. Supernatant was removed and cells were suspended in 1 ml of 75% ethyl alcohol. Cells were incubated at 4o C overnight. Following incubation, cells were centrifuged at 1000 rpm for 10 min. The supernatant was removed and cells were washed with 1 ml ice-cold PBS, pH 7.4. Cells were recentrifuged and re-diluted in 1 ml PBS pH 7.4. 0.25 μg/ml. RNase A was added and cells were incubated at 37o C for 30 min in order to remove any remaining traces of RNA that could interfere with PI. PI was added to a final concentration of 1 ug/ml. All experiments were performed in triplicate. The reported data constitute the average of three independent experiments.

RNA Isolation RNA isolation is one of the most crucial steps for the success of subsequent analyses since good RNA quality determines the yield in gene expression later studied with microarrays. In the present case the procedure followed was as follows: RNA was isolated with Trizol. The amount of RNA isolated was measured with a BioRad SmartSpec 3000 spectrophotometer and RNA integrity was estimated by 2% agarose gel electrophoresis. At least 40 μg of RNA from each sample was used. DNase treatment followed. Finally, RNA samples were further purified using a column-based kit and RNA amounts and integrity were determined again as above. Samples with a 1.8 to 2.0 A260/A280 ratio were selected. In addition, those RNA samples that empirically showed the 28S band twice as bright as the 18S band on the gel were utilized.

122

Microarray Experiments Since their advent, microarrays have been used to discover differences between biological samples. For example, on the level of gene expression, a sample is compared to a reference, in order to discover differences in the gene expression profile. Samples that are treated differently, for example a drug treatment against no treatment or normal cells against tumor cells, are used for RNA extraction and hybridized, either as cDNAs or as cRNAs (complementaryRNAs or also referred to as anti-sense amplified RNAs). Similarly, microarrays have been used for the detection of polymorphisms. During the past few years, microarray technologies have attracted a major deal of interest from the scientific community due to their potential in screening thousands of factors simultaneously. The first use of microarray technology was for detecting differences between two samples at the mRNA level. Microarrays consist of glass slides or membranes printed with oligonucleotides or cDNA fragments. Based on this property and the progress of high fidelity robotic technologies, microarrays have proven extremely valuable for the detection of literally tens of thousands of targets within one experiment. For this reason alone, this particular method has found applications in many areas of life sciences. With time, microarrays expanded to various types of high-throughput experimentations. Examples of these include methylation studies, SNP (Single Nucleotide Polymorphism) detection, CGH (Comparative Genomic Hybridization) microarrays and CHIP-chip (Chromatin Immunoprecipitation Chip) microarrays. A very important aspect of microarrays is the experimental design. Another important issue of microarray analysis is the normalization of data. This probably includes the most important aspect of microarray methodology, since it is the component that influences all further decisions and conclusions.


At this point we need to define the types of nucleic acids that are utilized in microarray analyses. The first and most abundantly used molecules were cDNAs. The in vitro reversely transcribed total RNA yields the respective cDNAs of all expressed exons at a certain time point under investigation. This method usually requires large amounts of total RNA as starting material, which ranges from 20-50ug per sample. Since this is not always the case, i.e. samples could be rare or difficult to obtain or of low quantity, other alternatives should be considered. A solution to these aforementioned problems came with the methodology of RNA amplification. It was first described by Van Gelder et al. in 1990 (Van Gelder et al., 1990). This methodology is based on the linear amplification of dsDNA transcribed to cRNA with use of the T4 DNA Polymerase, T3 RNA Polymerase and T7 RNA Polymerase. It is a very powerful technique since it can amplify 20ng to 2ug of total RNA yielding approximately a 5000-fold amplification. This means that samples of 200ng could yield 40ug of cRNA or samples of 2ug of total RNA could yield up to 100ug of cRNA. cRNA (complementary RNA) is also called aaRNA (anti-sense amplified RNA). Finally, when the desired molecules have been extracted and sample quality has been assured, nucleic acids are labeled with fluorochromes and let to hybridize. The basic and decisive step in microarray methodology is the sample selection and nucleic acid extraction, with RNAs or DNAs. In order to define the expression profile of genes at specific time points, one needs to “capture” the cell’s RNA or DNA at that specific time point. DNA and RNA quality defines the experimental output. The result is the ‘snap-shot” of the transcriptome at a given time point. Usually, in order to obtain more accurate information, it is necessary to perform multiple experiments at different time points or conditions, and then one speaks of a functional genomic assay. In the present case, for the assay of mRNA levels two sets of microarray chips were used:

cDNA microarray chips (4.8 kgenes) obtained from TAKARA (IntelliGene™ II Human CHIP 1) (Chung et al., 2004) and microarray chips (9.6k genes) from the Institut fuer Molekularbiologie und Tumorforschung, Microarray Core Facility of the Philipps-Universitaet, Marburg Germany (IMT9.6k). Hybridization was performed with the CyScribe Post-Labeling kit (RPN5660, Amersham) as described by the manufacturer. The fluorescent dyes used were Cy3 and Cy5. The RNA extracted form the CCRF-CEM cells was stained with Cy3 (reference/control) and RNA from the TE-671 cell line with Cy5 (experiment/ treatment). cDNAs were purified with QIAGEN PCR product clean-up kit (Cat # 28104). Slides were activated at 55o C for 30 min in 1% BSA. Samples were applied on the slides and let to hybridize overnight at 55o C. The following day, slides were washed in 200 ml 0.1× SSC and 0,1% SDS for 3× 5 min, in 200 ml 0.1× SSC for 2× 5 min and in 200 ml ddH2O for 30 sec. Slides were dried by centrifugation at 1500 rpm for 3 min and scanned with a microarray scanner (ScanArray 4000XL) (Perkin Elmer Inc. MA, former GSI Lumonics, USA). Images were generated with ScanArray microarray acquisition software (Perkin Elmer Inc. MA, former GSI Lumonics, USA).

Flow Cytometry Data Analysis Flow cytometry data contain a great amount of information that needs specialized software in order to be analyzed. There are several commercially available software packages and also several open-source software packages that are very good for flow cytometric data analysis. For the present analyses we have used the WinMDI software version 2.8 (The Scripps Institute, Flow Cytometry Core Facility http://facs.scripps. edu/software.html) and Cylchred version 1.0.2 (Cardiff University, Wales) which is based on the algorithms proposed by Watson et al. and Ormerod et al (Ormerod & Payne, 1987; Ormer-

123


od, Payne, & Watson, 1987; Watson, Chambers, & Smith, 1987).

Microarray Data Analysis The availability of software packages for microarray data analysis is really overwhelming. There are several commercially available software packages but there is also a very strong community of programmers and scientists that have created open-source software for microarrays. For the present analyses we have used mainly open-source software, except for the steps of image acquisitions and raw data extractions. Hence, microarray raw image analysis and raw data acquisition was performed with ImaGene® v.6.0 Software (BioDiscovery Inc, CA). ARMADA software (National Hellenic Research Foundation, Athens Greece, http://195.251.6.234/armada/) (Chatziioannou, Moulos, & Kolisis, 2009) was used for further filtering, normalization and clustering analyses. Normalization was performed using six different methods: a) No further processing after background correction, b) log2 transformation, c) LOWESS normalization, d) division with the Global Median (median of the 50% percentile), e) subtraction of Global Median (the median of the 50% percentile) and f) Rank Invariant with running Median. Finally, the Rank Invariant normalization method was chosen for further processing and analysis. Background correction was performed with global background correction, sub-grid based and negative control spots. Genes were filtered first for their quality of spot. This was done by acquiring a ‘0’ flag marking on each spot as assessed by ImaGene software (Signalto-Noise Ratio (SNR). Notice that Signal-to-Noise Ratio is calculated by the software from the equaµ − µB , where μR,G and μB is the tion SNR = R,G σB mean value intensity for the respective channel (Cy3 or Cy5) and mean background intensity respectively and σB is the background mean signal

124

standard deviation. A threshold of 2 has been set as a cut-off value, meaning that spot intensity for at least one channel should be twice as much than that of the background). Furthermore, each gene was tested for its significance in differential expression using a z-test. Genes were considered to be significantly differentially expressed if they obtained a p-value

Quality Assurance in Healthcare Service Delivery, Nursing and Personalized Medicine: Technologies and Processes

Service Design and Delivery

Quality Assurance

Integrative Neuroscience and Personalized Medicine

Dairy Processing and Quality Assurance

Dairy Processing and Quality Assurance

Healthcare, Medicine, and Science

Molecular Genetics and Personalized Medicine

Molecular Genetics and Personalized Medicine

International Practice Development in Nursing and Healthcare

Challenges in Analytical Quality Assurance

Challenges in Analytical Quality Assurance

Power Transformers Quality Assurance

Software Quality Assurance SOPs for Healthcare Manufacturers, Second Edition

Textbook of Personalized Medicine

Food Quality Assurance: Principles and Practices

Numeracy in nursing and healthcare : calculations and practice

Welding: Processes, Quality, and Applications

Study Skills for Nursing and Healthcare Students

Essentials of Genomic and Personalized Medicine

Research Methods for Nursing and Healthcare

Research Methods for Nursing and Healthcare

Software Quality Engineering: Testing, Quality Assurance, and Quantifiable Improvement

Software Quality Engineering: Testing, Quality Assurance, and Quantifiable Improvement

Software Quality Engineering: Testing, Quality Assurance, and Quantifiable Improvement

Assisted Reproductive Technologies Quality and Safety: Quality and Safety

Developments in Healthcare Information Systems and Technologies: Models and Methods

RNA Technologies in Cardiovascular Medicine and Research

Information Technologies in Medicine

Quality Assurance in Environmental Monitoring: Instrumental Methods

Quality Assurance in Dialysis, Second Edition

Quality Assurance in Healthcare Service Delivery, Nursing and Personalized Medicine: Technologies and Processes

Service Design and Delivery